Keywords: Biological systems

Keywords: Intelligent systems

Keywords: Chemical process control, Machine learning, Control applications
Abstract: This work presents a reinforcement learning (RL) based methodology for generating explicit control laws for nonlinear systems with input constraints. By utilizing the Deep Deterministic Policy Gradient (DDPG) algorithm, we eliminate the need for prior dataset collection and knowledge about the model of the system while producing explicit control laws. We propose transformations and offer specific setting choices to make the DDPG algorithm generate an explicit control law with almost the same performance and behavior as Nonlinear Model Predictive Control (NMPC).

Explicit control laws are essential for real-time control applications, especially in embedded environments with limited computational resources. Although explicit solutions for linear systems have been thoroughly examined, extending them to nonlinear systems still poses a challenge. Our approach produces explicit control laws that maintain near-NMPC performance. The resulting control law, designed as a neural network, requires minimal storage, making it suitable for implementation on microcontrollers.

The methodology is validated through a case study involving a continuous stirred tank reactor (CSTR). Our DDPG framework's control law achieves performance comparable to NMPC without the necessity of collecting a dataset or gaining knowledge of a model. Despite being trained in a limited disturbance scenario, the RL-based explicit control law successfully generalizes to unseen disturbances, demonstrating its robustness. Performance comparisons indicate that the RL-generated explicit control law results in only a (3.84%) reduction in control performance compared to an NMPC benchmark while maintaining feasibility and adherence to constraints. These results underscore reinforcement learning's potential for real-time control applications.

We acknowledge the contribution of the Scientific Grant Agency of the Slovak Republic under the grants VEGA 1/0239/24. P. Valábek is also supported by an internal STU grants for young researchers. M. Klaučo is also supported by the European Union project ROBOPROX (Reg. No. CZ.02.01.01/00/22_008/0004590).

Keywords: Optimization, Optimization algorithms, Optimal control
Abstract: Solving real-time quadratic programming (QP) is a ubiquitous task in control engineering, such as in model predictive control and control barrier function-based QP. In such real-time scenarios, certifying that the employed QP algorithm can either return a solution within a predefined level of optimality or detect QP infeasibility before the predefined sampling time is a pressing requirement. This article considers convex QP (including linear programming) and adopts its homogeneous formulation to achieve infeasibility detection. Exploiting this homogeneous formulation, this article proposes a novel infeasible interior-point method (IPM) algorithm with the best theoretical O(sqrt{n}) iteration complexity that feasible IPM algorithms enjoy. The iteration complexity is proved to be textit{exact} (rather than an upper bound), textit{simple to calculate}, and textit{data independent}, with the value leftlceilfrac{log(frac{n+1}{epsilon})}{-log(1-frac {0.414213}{sqrt{n+1}})}rightrceil (where n and epsilon denote the number of constraints and the predefined optimality level, respectively), making it appealing to certify the execution time of online time-varying convex QPs. The proposed algorithm is simple to implement without requiring a line search procedure (uses the full Newton step), and its C-code implementation (offering MATLAB, Julia, and Python interfaces) and numerical examples are publicly available at url{https://github.com/liangwu2019/EIQP}.

Keywords: Cooperative control, Adaptive control, Energy systems
Abstract: We investigate the distributed state-of-charge (SoC) balancing control problem of networked heterogeneous battery systems with unknown battery parameters in the presence of actuator attacks. Existing distributed SoC balancing control algorithms often rely on assumptions such as having homogeneous or known battery parameters, operating under undirected communication topologies and knowing global information of the communication topology, and lack resilience against actuator attacks. In this study, we aim to relax these assumptions by developing a fully distributed adaptive resilient control algorithm for each battery unit. It is shown that under a strongly connected directed communication topology, the control objectives of SoC balancing and proportional power sharing are achieved among battery units with heterogeneous and unknown battery parameters without using any global information, even in the presence of actuator attacks. Through simulation results in MATLAB/Simulink, we validate the effectiveness of the proposed control algorithm in discharging or charging mode and highlight its superiority in enhancing resilience against actuator attacks.

Keywords: Neural networks, Pattern recognition and classification, Stability of nonlinear systems
Abstract: Oscillatory Ising machines (OIM) are actively being investigated as alternate compute engines to efficiently solve large-scale combinatorial optimization problems, many of which are NP-hard problems. Many experimental OIM prototypes show a common phenomenon wherein phases of coupled oscillators bifurcate and converge to either 0 or π if the effect of sub-harmonic injection locking is sufficiently strong. As a dynamic system, this phenomenon, roughly speaking, is equivalent to the asymptotic stability property of an equilibrium point of the OIM. In this work, we provide an in-depth analysis of the dynamic properties of all-to-all connected OIMs from a control theoretic point of view and explore their application as associative memory.

Keywords: Modeling, Energy systems, Neural networks
Abstract: This study presents a novel Graph Neural Network (GNN)-based surrogate model for predicting state evolution in reconfigurable battery packs. By leveraging graph-based representations of battery cell interconnections, the proposed approach addresses the unique challenge of estimating the imbalance in state-of-charge (SOC) and temperature of cells of a battery pack in dynamic battery configurations. Unlike conventional methods that focus on instantaneous state estimation, our GNN model predicts future SOC and temperature distributions by considering both current system state and switch configuration. The model architecture combines Graph Attention Networks with pooling operations to effectively capture cell-to-cell interactions and battery pack-level dynamics. Numerical results demonstrate that our GNN-based approach significantly outperforms baseline Feedforward Neural Network (FNN) and FNN-attention models, showing substantial improvements in prediction accuracy for both temperature and SOC while maintaining robust performance even with limited training data.

Keywords: Machine learning, Optimal control
Abstract: In this research, we propose an Inverse Reinforcement Learning (IRL)-based Model Predictive Control (MPC) framework to achieve personalized lane-change maneuvers for autonomous vehicles. By analyzing expert human driving data, we define interpretable driving features—including lateral offset, heading error, yaw rate, and steering angle—which are used in IRL to infer personalized driving styles. The resulting cost function is directly integrated into an MPC controller, allowing it to generate control commands that emulate human-like lane-change behaviors. Experimental validations conducted in the CARLA simulator confirm that our method successfully reproduces smooth and stable lane changes closely aligned with expert demonstrations, highlighting the potential for adaptive and personalized autonomous driving

Keywords: Predictive control for nonlinear systems, Numerical algorithms, Kalman filtering
Abstract: Model predictive control (MPC) is a powerful optimization-based control framework, but its application to high-dimensional nonlinear systems, such as numerical weather prediction (NWP) models, is often limited by computational costs. To address this challenge, we propose ensemble model predictive control (EnMPC), a novel method that integrates MPC with ensemble data assimilation to reduce the cost of full model evaluations. By approximating the MPC cost function using ensembles, EnMPC effectively accounts for nonlinearity and uncertainty while improving computational efficiency. We applied EnMPC to an atmospheric model (SCALE) and demonstrated its potential through a case study of a severe rainfall event in eastern Japan in September 2015. The control objective was to steer the model trajectory toward an ensemble member with minimal precipitation. The control inputs were small perturbations applied to the model’s initial conditions within a forecast window. Our experiments showed that EnMPC can influence the evolution of atmospheric fields toward desired outcomes with reduced rainfall, indicating the feasibility of controlling NWP systems in practice. This framework provides a new pathway for efficient control in high-dimensional nonlinear systems and may be extended to other domains where uncertainty-aware decision-making is required.

Keywords: Optimal control, Energy systems, Simulation
Abstract: Electricity is essential for daily life, yet disasters can cause outages, requiring alternative solutions. This study proposes an EV-based electricity delivery system for disaster-affected households by predicting electricity demand and optimizing EV travel routes to ensure efficient distribution. The study assumes a scenario where households with photovoltaic systems and storage batteries are disconnected from the power grid. Multiple EVs and charging stations are available. Household electricity demand is predicted using historical consumption data and temperature as an explanatory variable, applying Long Short-Term Memory for forecasting. EVs prioritize households based on battery state-of-charge thresholds. The EVs prioritize households based on battery state-of-charge thresholds. Two optimization algorithms are tested for route planning of the EVs. The first clusters households into groups, assigns each EV a group, and generates a fixed circulation route to minimize total travel distance while balancing energy load using the 2-opt method and Tabu search. The second dynamically assigns the nearest unserved household to an EV in real time, minimizing distance per trip. At each step, EVs determine whether recharging is necessary before reaching the next household. If required, they recharge at the nearest station before continuing distribution. A computer simulation evaluates the proposed method using real electricity consumption data across different numbers of EVs, household counts, map sizes, and seasonal periods. Results show that with fewer EVs, the first algorithm performs better by reducing unnecessary movement, while with more EVs, the second algorithm is more efficient by dynamically optimizing short-term routing. Over longer durations, the first algorithm is advantageous due to its pre-determined patrol routes, whereas the second algorithm provides greater flexibility in short-term operations. The effectiveness of each algorithm depends on map size, EV availability, and time duration. The first algorithm is preferable for long-term efficiency, while the second is better suited for short-term flexibility. Future work will explore hybrid models integrating both approaches for improved performance.

Keywords: Optimization algorithms, Stability of nonlinear systems, Autonomous robots
Abstract: Controllers which solve optimization problems in real-time are prevalent. Examples of this paradigm include model predictive control, control barrier function-based controllers, and online feedback optimization. However, the burden of repeatedly solving the parametrized optimization problem in real-time may be overwhelming in some applications. In this poster, we showcase how we can use contracting dynamics with a feedforward correction term to ensure exact tracking of the solution of the parametrized optimization problem while only needing to simulate an auxiliary dynamical system. To illustrate our approach, we present a use-case in a multi-robot collision avoidance problem in the Robotarium.

Keywords: PID control, Energy systems, Control applications
Abstract: Floating Offshore Wind Turbines (FOWTs) are pivotal to accessing deep water wind resources near population centers. However, traditional fixed-bottom wind turbine blade pitch controllers usually lead to adverse platform pitch motion when applied “as is” to FOWTs. Feeding back the platform motion to the blade pitch or the generator torque controller commands are two common methods for combatting the platform’s instability in the loop with land-based wind turbine controllers. A largely successful blade-pitch based platform feedback has been previously implemented on NREL’s Reference Open-Source Controller (ROSCO). This work extends the prior work by demonstrating that use of a generator-torque based platform feedback enables re-tuning of the baseline pitch controller for the IEA 15 MW Reference Wind Turbine. In particular, we use the ROSCO platform to implement a torque-based floating feedback controller in parallel with several re-tuned versions of the baseline pitch controller. Simulations of the plant and the controllers were done on the OpenFAST platform with the ROSCO control architecture. We show that faster bandwidths can be enabled when torque control is used to reduce platform motion than when the pitch control loop is used, with corresponding benefits to several signals of interest including generator speed, and platform pitching. Observations were analyzed and results indicate that at the conditions tested, both approaches possess comparative advantages. While the torque-based approach accrued more counts of comparative advantages than the pitch-based approach, the ultimate decision about which to use comes down to designer’s choice related to the advantages and disadvantages of each.

Keywords: Neural networks, Autonomous robots, Reinforcement learning
Abstract: Neural network-based policies have demonstrated success in many robotic applications, but often lack human-interpretability, which poses challenges in safety-critical deployments. To address this, we propose a neuro-symbolic explanation framework that generates a weighted signal temporal logic (wSTL) specification to describe a robot policy in a human-interpretable form. Existing methods typically produce explanations that are verbose and inconsistent, which hinders explainability, and loose, which do not give meaningful insights into the underlying policy. We address these issues by introducing a simplification process consisting of predicate filtering, regularization, and iterative pruning. We also introduce three novel explainability evaluation metrics---conciseness, consistency, and strictness---to assess explanation quality beyond conventional classification metrics. Our method is validated in three simulated robotic environments, where it outperforms baselines in generating concise, consistent, and strict wSTL explanations without sacrificing classification accuracy. This work bridges neural network policy learning with formal methods, contributing to safer and more transparent decision-making in robotics.

Keywords: Human-in-the-loop control, Networked control systems, Linear systems
Abstract: Conventional event-triggering frameworks typically depend on explicit knowledge of continuous-time or discrete-time control law structures. However, not all control laws have an explicitly defined structure, particularly those generated by human operators, machine learning algorithms, and/or computational optimization methods. Motivated by this limitation, we seek to fill this critical scientific gap by proposing a novel event-triggering framework that is independent of any particular control law structure and solely utilizes open-loop system dynamics for scheduling control and state data transmissions. In addition to the proposed framework, an illustrative numerical example involving a human subject is provided to demonstrate the efficacy of our contribution.

Keywords: Machine learning
Abstract: New challenges arise from the urge of bringing machine learning (ML) in safety-critical environments as safe operations have to be ensured. However, ML models cannot be fully tested yet. To cope with this weakness Uncertainty Quantification (UQ) is utilized to quantify the certainty of the model. In UQ, the current focus is on data driven approaches like ensembles and Bayesian neural networks. Nonetheless, these approaches cannot estimate the uncertainty in unseen areas, raising the same issue as for the ML model. Therefore, they cannot be used for certification purposes within safety-critical domains. This work proposes an early-stage conceptual holistic framework for UQ called Lifecycle Uncertainty Quantification in ML (UQML). Our approach calculates the uncertainty emerged during training, testing and operation unlike the state-of-the-art data driven methods. The poster introduced relevant UQ methods used in the entire lifecycle, such as, out-of-distribution detection, exploratory data analysis, training weight change evaluation, model analysis, feature evaluation and impact quantification, formal method testing, ODD evaluation and out of ODD detection, distance to seen data and ontology based error detection among others. The proposed framework is illustrated through a runway detection use case.

Keywords: Control of networks, Model/Controller reduction, Computational methods
Abstract: We present a coupled PDE model for a clamped-free transmission line comprising two interconnected strings and a dynamic interior mass. Earlier works employed lower-order velocity-based feedback—either at the joint (Chen-Coleman-West, 1987; Lee-You, 1989) or the right-end boundary (Hansen-Zuazua, 1995; Littman-Taylor, 2002)—but these designs often fail to ensure exponential stability. To address this, we introduce a higher-order feedback sensor incorporating angular velocity at the joint. Building on (Morgul-Rao-Conrad, 1994), we use a Lyapunov-based approach to prove exponential stability with an explicit decay rate. The analysis extends to a structure-preserving Finite Difference scheme, showing that the semi-discrete system inherits the decay rate of the PDE model. We also provide theoretical results on exponential and polynomial stability under partial controller deactivation.

Keywords: Neural networks, Robotics, Autonomous robots
Abstract: This poster focuses on driving a differential wheel robot with low computing power neural networks on board by distributing neurons in need. Our approach is to find the minimal use of neural and adaptive MLP structure. In our application we will use an Elisa-3 robot to navigate through an unknow path. As the robot explores the environment it will organize the predefined neural architecture inside the system. The goal is to drive the robot get an optimal turning angle and turning speed. Network architecture will be refined responsively using Ant Colony Optimization (ACO) to activate or deactivate neural connections based on use. The use of one side of the robot’s actuators more than the others will stimulate the relevant neural connections while unstimulated connections decay in weight based on the ACO pheromone model. The more a network region is used the more neurons will be added to the structure to give it more processing power. In the testing case of circling a track, for each round of the robot circling through the path the training data will be collected, then ACO will be used to evolve the network structure.

Keywords: Nonlinear systems identification, Predictive control for nonlinear systems, Markov processes
Abstract: Data-driven Koopman-theoretic approaches have proven effective in output prediction, state estimation, and control of nonlinear dynamical systems. For control-affine systems, the Koopman generator's affine input dependence enables finite-dimensional bilinear approximations. However, a significant challenge in constructing Koopman generators for actuated systems is its reliance on appropriate basis functions or observables, with no unified framework for their selection. Real-world applications often involve noisy, partially observed states, requiring Koopman observables to capture system behavior from input-output data. The challenge of identifying Koopman observables under partial observation is well studied, e.g., time-delayed observables offering viable solutions. However, the presence of actuation reduces the efficacy of these methods. To address this limitation, a recent data-driven method leverages a learned Koopman generator-based bilinear surrogate model with linear reconstruction, demonstrating promise for actuated nonlinear system identification. Further study is needed to assess its effectiveness in complex, partially observed nonlinear systems with actuation and sensitivity to initialization. To address this, we model the dynamics of a control-affine nonlinear system as a bilinear Hidden Markov Model (HMM) defined via Koopman generators with a nonlinear observation map (decoder) modeled using a multilayer perceptron (MLP). The parameters of the HMM and decoder are learned from noisy output data using a neural expectation maximization (EM) approach. In the EM method, the E-step employs an extended Kalman filter and smoother, while the M-step utilizes a least-squares approximation of the Koopman generators, combined with a gradient-based optimization for the decoder parameters. In addition, we present a model-predictive control (MPC)-based output regulation method using the learned HMM as a predictive model. We demonstrate the performance of our method on three nonlinear systems: (1) an actuated polynomial system with a slow manifold and partial observation, (2) a forced Duffing oscillator with partial observation, and (3) an unforced Kuramoto-Sivashinsky equation with noisy observation.

Keywords: Emerging control applications, Reinforcement learning, Predictive control for linear systems
Abstract: Various chronic health conditions, including cardiovascular disease, breast and colon cancer, obesity, diabetes, and arthritis, are associated with insufficient levels of physical activity. These diseases reduce the quality of life of patients and may have fatal consequences. Physical activity (PA) can serve as both prevention and as treatment; recent studies have shown that walking an average of 8,000 steps/day (from 4,000 steps/day) can reduce the risk of these conditions by 51%. This poster considers a currently ongoing mobile health (mHealth) PA intervention YourMove, a first-of-its-kind control optimization trial (COT) study. Grounded in Social Cognitive Theory (SCT), YourMove relies on model predictive control (MPC) to deliver daily physical activity goals and tailored text messages via a smartwatch or fitness tracker, encouraging adults to engage in moderate-to-vigorous physical activity. MPC simulated results demonstrate that adherent participants achieve targeted step goals and improved self-efficacy despite external disturbances like temperature fluctuations. Building on this approach, our NSF-funded ExpandAI project seeks to incorporate reinforcement learning (RL) into the YourMove intervention by designing an aligned reward function that integrates a participant’s expressed preferences for enhanced personalization. The methodology considers partially observable factors like psychosocial contexts, using Bayesian inference to incorporate explicit and implicit user feedback into the reward function. Based on utility theory, the reward structure assigns utility scores to intervention trajectories based on stakeholder preferences. Future work will evaluate various aligned reward functions within combined control and RL frameworks, for which RL parameters (states, observations, actions, rewards) have been defined for both open-loop and closed-loop scenarios. This will refine YourMove, targeting robust, personalized behavior change and increased user engagement.

Keywords: Nonlinear systems identification, Statistical learning, Grey-box modeling
Abstract: Nonlinear autoregressions, including recent SINDy and Koopman operator interpretations, are biased in the presence of measurement noise. This problem was first appreciated in the 1970s in the context of State Variable Filtering (SVF), which identified continuous-time linear systems by least-squares regressions of numerical derivatives. In the 2010s, similar regressions constitute the final step (following model structure selection) when learning nonlinear dynamics from time series data. We revisit the "original sin" of SVF, namely, asymptotic bias that worsens with measurement noise. We reinvent two bias reduction techniques—bias correction and instrumental variables—for continuous-time nonlinear systems. Benchmarked against denoised least squares on a replication of the SINDy Lorenz system, we achieve up to 3000x reduction in parameter bias and 3x reduction in parameter RMSE.

Keywords: Autonomous robots, Optimal control, Uncertain systems
Abstract: We present a novel framework for robotic source seeking that enables mobile robots to efficiently locate sources in complex and uncertain environments. Our approach leverages Behavioral Entropy (BE), a generalized uncertainty measure, to systematically encode a range of navigation behaviors during source seeking. In cases where the signal strength is low, we incorporate a Gaussian Process(GP)- based exploration strategy to gather additional information and reduce map uncertainty. Our framework operates in the continuous domain, integrating Wasserstein gradients of BE and differential entropy gradient of GP to guide the robot towards regions with stronger signals. Furthermore, we develop an adaptive tuning algorithm that dynamically adjusts the BE parameters, allowing the robot to smoothly transition between aggressive exploitation and cautious exploration based on real-time conditions. Simulation results in ROS-Unity environments demonstrate the effectiveness of our method and its advantages over traditional approaches that rely on fixed uncertainty measures like Shannon entropy.

Keywords: Numerical algorithms, Reinforcement learning, Optimal control
Abstract: This poster presents a novel gradient-based method for solving the linear quadratic regulator (LQR) problem for linear systems with nonlinear dependence on time-invariant probabilistic parametric uncertainties. The method explicitly addresses uncertainty in the model parameters, ensuring robust performance. The approach leverages polynomial chaos theory (PCT) alongside advanced policy optimization (PO) techniques to develop a gradient method that directly optimizes the static state-feedback gain. Specifically, by applying PCT, the original stochastic system is expanded into a high-dimensional linear time-invariant (LTI) system with structured state-feedback control. A first-order gradient descent algorithm is then introduced to update the structured state-feedback gain, iteratively minimizing the LQR cost. Numerical examples highlight the superior computational efficiency of this approach compared to conventional bilinear matrix inequality (BMI)-based methods.

Keywords: Optimization algorithms, Smart grid, Power systems
Abstract: We consider an optimal flow distribution problem in electric power systems in which the goal is to find a radial configuration that minimizes resistance-induced quadratic distribution costs while ensuring delivery of inputs from multiple sources to all sinks to meet their demands. This problem has critical effects over society, where efficient energy flow is crucial for both economic and environmental reasons. Due to its complexity, finding an optimal solution is computationally challenging and NP-hard. In this paper, we study the effectiveness of FORWARD algorithm, which leverages graph theory to efficiently identify feasible configurations in polynomial time. By drawing parallels with random walk processes, our method simplifies the search space, significantly reducing computational effort while maintaining performance. The FORWARD algorithm employs a combination of network preprocessing, intelligent partitioning, and strategic sampling to construct radial configurations that meet flow requirements, finding always a feasible solution. Numerical experiments demonstrate its effectiveness, highlighting its potential for real-world applications in optimizing distribution networks.

Keywords: Machine learning, Optimization, Adaptive control
Abstract: Endowing Reinforcement Learning algorithms with model information to control Hamiltonian systems has been shown to improve convergence and learning of the control policies. Furthermore, recent works in deep learning have introduced ways to include physics-informed inductive biases in learning interpretable Hamiltonian dynamics. Building upon these works, we develop an end-to-end iterative learning framework that learns an optimal energy-based control from observed real-system trajectory data. The proposed framework incorporates the structure of underlying system dynamics and associated optimal policy parameterized by neural networks to improve the interpretability of the learned optimal policy. The learned policy is performance-based and renders the system inherently passive and provably stable under minimal assumptions. The proposed approach is validated on state-regulation tasks.

Keywords: Network analysis and control, Communication networks, Control of networks
Abstract: This work quantifies the impact of network topology on the opinions of agents formed by social interaction. In this work, we consider that the opinions of agents evolve by the Friedkin-Johnsen model with heterogeneous agents showing varying degrees of stubbornness towards interactions. We consider the underlying network to form a signed directed acyclic graph which represents the hierarchical interactions in political, military, or business organisations. In this framework, two types of influential agents exist, distinguished by their topological position and stubborn behaviour. We propose the construction of a signal flow graph (SFG) from the underlying interaction network to illustrate the role played by the network in shaping the final opinions of the agents. We demonstrate that the overall influence of an influential agent depends on the paths originating from it to the rest of the graph. Importantly, we show that in signed digraphs, the influential agents are not always competing in a zero-sum game like in cooperative networks. This analysis finds application in examining the impact of recommender systems on the opinions of individuals that alter social interactions to provide biased and sponsored content with the objective of maximising engagement and revenue.

Keywords: Neural networks, Machine learning, Pattern recognition and classification
Abstract: Modern deep neural networks often require significant depth and large parameter counts to approximate complex functions. COPNet introduces a recursive, mathematically grounded architecture that achieves efficient approximation with fewer parameters and more robust training dynamics. The Compositional Orthogonal Polynomial Neural Network (COPNet) constructs hierarchical representations using orthogonal polynomial expansions. Unlike traditional networks with independent layer activations, COPNet applies a modified three-term recurrence relation: P_(n+1) (x)=(A_n x+B_n ) P_n (x)-C_n P_(n-1) (x), for n=0,1,2,… Where P_(-1)=0, and P_0=1. If the leading coefficient of P_n (x) is k_n>0, and A_n, B_n, and C_n are constants, A_n=k_(n+1)/k_n , C_(n+1)=A_(n+1)/A_n h_(n+1)/h_n . Each layer builds upon the two previous layers, incorporating a trainable transformation. This recursive framework enables effective function approximation, reduces feature redundancy, and maintains well-behaved gradient propagation across depth. COPNet was evaluated on two-dimensional sinusoidal function approximation tasks. Results show faster convergence and lower validation loss compared to ReLU, Sigmoid, and Tanh-based networks, using roughly 50% fewer parameters. These results highlight COPNet’s promise as an efficient and stable alternative to conventional deep architectures.

Keywords: Adaptive control
Abstract: This study presents a novel automated fluid resuscitation framework designed to maintain hemodynamic stability in the presence of limited and noisy physiological data. We propose a robust nonlinear state-space modeling (RNSSM) algorithm, trained via variational autoencoder learning, to capture mean arterial pressure (MAP) responses to fluid infusion in hemorrhagic scenarios. The model is integrated with a radial basis function (RBF) optimal control approach that combines function approximation and predictive optimization to regulate fluid infusion dosages during resuscitation. The accuracy of the RNSSM was confirmed using real-world data. Additionally, the superior performance of the RBF optimal controller was demonstrated in comparison with state-of-the-art fluid resuscitation control algorithm. Simulation results indicate that this approach addresses key limitations of existing methods by enabling more accurate, subject-specific hemodynamic regulation for fluid management in critical care.

Keywords: Control applications, Intelligent systems, Manufacturing systems
Abstract: Current robot autonomy struggles to operate beyond the assumed Operational Design Domain (ODD), the specific set of conditions and environments in which the system is designed to function, while the real-world is rife with uncertainties that may lead to failures. Automating recovery remains a significant challenge. Traditional methods often rely on human intervention to manually address failures or require exhaustive enumeration of failure cases and the design of specific recovery policies for each scenario, both of which are labor-intensive. Foundational Vision-Language Models (VLMs), which demonstrate remarkable common-sense generalization and reasoning capabilities, have broader, potentially unbounded ODDs. However, limitations in spatial reasoning continue to be a common challenge for many VLMs when applied to robot control and motion-level error recovery. In this paper, we investigate how optimizing visual and text prompts can enhance the spatial reasoning of VLMs, enabling them to function effectively as black-box controllers for both motion-level position correction and task-level recovery from unknown failures. Specifically, the optimizations include identifying key visual elements in visual prompts, highlighting these elements in text prompts for querying, and decomposing the reasoning process for failure detection and control generation. In experiments, prompt optimizations significantly outperform pre-trained Vision-Language-Action Models in correcting motion-level position errors and improve accuracy by 65.78% compared to VLMs with unoptimized prompts. Additionally, for task-level failures, optimized prompts enhanced the success rate by 5.8%, 5.8%, and 7.5% in VLMs' abilities to detect failures, analyze issues, and generate recovery plans, respectively, across a wide range of unknown errors in Lego assembly.

Keywords: Robotics, Autonomous robots, Mechanical systems/robotics
Abstract: This work proposes a nonlinear motion control for a multi-drone slung load system (MSLS), which consists of two multirotor drones carrying a shared slung load modeled as a point mass suspended by rods. The control objective is time-varying tracking of a six-dimensional trajectory consisting of payload position and drone-payload shape. A novel input transformation and flat output are proposed, essential for achieving exact dynamic state feedback linearization. The resulting tracking error dynamics is linear-time-invariant, and exponential stability is guaranteed on a practical subset of state space. The publicly available design is validated with open-source software-in-the-loop (SITL) simulation.

Keywords: Autonomous systems, Aerospace
Abstract: This paper presents a quadrotor guidance method for visual inspection from specific viewpoints along an elliptic boundary surrounding a structure. The proposed guidance method leverages a new bifurcation theory-based approach to achieve two inspection modes: reaching a desired viewpoint and switching between viewpoints along the elliptical boundary. The stability analysis confirms that the quadrotor achieves the inspection modes as the bifurcation parameter varies. Numerical simulations illustrate the effectiveness of the proposed method.

Keywords: Robotics, Networked control systems, Agents-based systems
Abstract: Extensive research on graph-based dynamics and control of multi-agent systems has successfully demonstrated control of robotic swarms, where each robot is perceived as an independent agent virtually connected by a network topology. The strong advantage of the network control structure lies in the decentralized nature of the control action, which only requires the knowledge of virtually connected agents. In this paper, we seek to expand the ideas of virtual network constraints to physical constraints on a class of tree-structured robots which we denote as single articulated robotic (SAR) systems. In our proposed framework, each link can be viewed as an agent, and each holonomic constraint connecting links serves as an edge. By following the first principles of Lagrangian dynamics, we derive a consensus-like matrix-differential equation with weighted graph and edge Laplacians for the dynamics of a SAR system. The sufficient condition for the holonomic constraint forces becoming independent to the control inputs is derived. This condition leads to a decentralized leader-follower network control framework for regulating the relative configuration of the robot. Simulation results demonstrate the effectiveness of the proposed control method.

Keywords: Robotics, Optimization
Abstract: Geodesic path planning is crucial in applications such as robotics, computer graphics, and autonomous navigation, focusing on finding the shortest path between two points on a curved surface while accounting for intrinsic geometry. Traditional methods, including energy function minimization, heat flow, and curvature-based techniques, often face local minima and computational inefficiencies, particularly in irregular environments. This paper presents a novel method that integrates the exponential map with the injectivity radius, ensuring globally optimal paths. Our approach avoids local minima, guarantees the shortest path, and provides real-time computational efficiency. Simulations show that our method outperforms existing techniques in optimality, local minima avoidance, and computation time.

Keywords: Multivehicle systems, Autonomous robots, Robotics
Abstract: We introduce FlySurf, a novel flying robotic surface that integrates shape morphing control to achieve enhanced maneuverability and adaptability in dynamic environments. FlySurf is modeled as a mesh-based dynamic structure, allowing for complex deformations and flexibility during flight. This approach represents a significant contribution to aerial robotics, where achieving simultaneous flight control and shape adaptation remains a challenging task. For efficient shape morphing control, we propose a linear quadratic Gaussian (LQG) control strategy that integrates a linear quadratic regulator (LQR) with an extended Kalman filter (EKF) for state estimation from limited observations. We validate the proposed FlySurf framework through a series of numerical simulations, demonstrating its ability to perform maneuvers while adapting its shape.

Keywords: Emerging control applications, Hybrid systems, Identification for control
Abstract: This paper describes a systematic approach for epidemic control using control-relevant identification coupled with a multi-input, multi-output 3-Degree-of-Freedom Kalman filter-based Hybrid Model Predictive Control (MIMO 3DoF-KF HMPC) featuring online controller reconfiguration. The combined data-driven modeling and control strategy is evaluated on a Susceptible-Infected-Recovered (SIR) model involving vaccination and loss of immunity (i.e., reinfection). "Zippered" multisine input signals and ARX estimation are applied to obtain a multivariable dynamic model that is the basis for an HMPC algorithm featuring both continuous and categorical (i.e., discrete level) actions through a Mixed Integer Quadratic Programming (MIQP) formulation. The goal is to reduce the infected population while balancing societal impacts. The hybrid formulation with reconfiguration dynamically adjusts health intervention policies, such as categorical lockdown levels and continuous vaccination rates. This approach enables operational goals such as reducing the infected population to a desired interval or relaxing lockdown to a designated setpoint; furthermore, the 3DoF formulation enables independent tuning for setpoint tracking and measured and unmeasured disturbance rejection, allowing scalable solutions across diverse epidemiological settings. The framework is demonstrated through two demanding case studies involving 90% infection reduction under time-varying recovery and loss of immunity. The resulting closed-loop model provides a practical tool for guiding government policy and public decisions during pandemics.

Keywords: Biomedical, Modeling, Identification
Abstract: Type 1 Diabetes Mellitus (T1DM) is an autoimmune condition characterized by the destruction of pancreatic beta cells, leading to insulin deficiency and requiring lifelong exogenous insulin administration. Effective management of T1DM depends on accurate insulin dosing, a challenge due to the dynamic nature of glucose-insulin interactions, which varies both between and within individuals over time due to variations in insulin sensitivity (SI ). This paper presents an extension of the physiological long-term glucose-insulin model initially proposed by Ruan et al. [1], incorporating a novel nonlinearity in it. The new model reflects SI variability as a function of both physiological variables and circadian rhythms. By capturing the temporal fluctuations in SI, the model aims to enhance the predictive capability of glucose-insulin models without increasing complexity, facilitating future integration into real-time control systems like model predictive control (MPC) in artificial pancreas (AP) systems. Simulation scenarios with the UVA/Padova T1DM simulator validate the model, demonstrating improvements in blood glucose modeling compared to existing methods.

Keywords: Biomedical, Hybrid systems, Optimization
Abstract: Shock is a life-threatening condition that requires immediate hemodynamic stabilization. While blood pressure management is prioritized, maintaining adequate blood volume and perfusion has been associated with improved outcomes. There are two major challenges to automating hemodynamic control in a clinical practice environment such as the ICU. (1) Measurement limitations. While blood pressure can be measured continuously, cardiac output and left atrial pressure must be measured at discrete intervals via a pulmonary arterial catheter. (2) Control system complexity. Cardiovascular response to multiple drug infusions is highly nonlinear and interdependent, such that hemodynamic control of multiple targets via multiple drugs has been challenging. In this paper, we propose a hybrid hemodynamic control system, where cardiac output and left atrial pressure are discretely controlled while blood pressure is controlled continuously and preferentially. After each discrete measurement, an optimization problem based on a pharmacological effects model and circulatory equilibrium theory is solved to identify an optimal drug composite that achieves the desired hemodynamic control targets. Between discrete measurements, blood pressure is continuously controlled by adjusting the amplitude of the identified drug composite. Simulation studies were conducted for (a) hypovolemic, (b) septic, and (c) cardiogenic shock using a high-fidelity hemodynamic simulator. The results demonstrate successful continuous control of blood pressure, with discrete improvements in cardiac output and left atrial pressure control. This application is clinically applicable because it explicitly takes into account the limitations of a clinical measurement environment.

Keywords: Biomedical, Cellular dynamics, Game theory
Abstract: Mechanical ventilation is the primary treatment to support lung function during acute respiratory distress syndrome (ARDS), a form of severe respiratory dysfunction. Mechanical ventilation protocols can vary significantly across hospital systems, and the variability can impact the quality of care. Up to 40% of patients with ARDS develop multiple organ dysfunction syndrome, with patient mortality ranging from 30-50% dependent on ARDS severity. A control algorithm for mechanical ventilation was developed and applied to a physiologically-motivated model for oxygen, waste, and inflammation trafficking that integrates a game theoretic approach for cellular decision-making. A pathogen insult within the lung acts as a disturbance decreasing oxygenation and triggering a systemic response, which leads to controller activation as lung function declines. Mechanical ventilation settings are adjusted along clinically-validated protocols to maintain oxygen saturation within desired bounds, thereby prolonging survival and allowing recovery from ARDS. Two mechanical ventilation protocols are compared in their respective improvement in virtual patient outcome, as measured by recovery, survival, or mortality. Both protocols produced an equivalent increase in recovery across a two-dimensional parameter space, which is consistent with clinical trial outcomes.

Keywords: Biomolecular systems, Cellular dynamics, Biomedical
Abstract: We propose an enhanced biomolecular controller that detects and treats gut infection utilizing two treatment actuation strategies: quorum quenching and pyocin-mediated killing. The quorum quenching module disrupts bacterial communication signals, reducing virulence and antibiotic resistance of the pathogen that causes the infection. The pyocin module specifically targets and eliminates this pathogen. We model and simulate theoretical performance under varying pathogen dynamics by integrating new treatment strategies into our existing detector-population controller framework, establishing a unified predictive model that more accurately captures the treatment process. These simulations, which incorporate medically relevant dynamics, provide deeper insights into design requirements and contribute to the development of engineered cell therapies, demonstrating the potential of a robust and automated response to infections. More broadly, our approach demonstrates how theoretical modeling and rigorous analysis serve as foundational steps in uncovering underlying design principles, enabling the rational development of therapeutic strategies that are both effective and adaptable across different biomedical applications.

Keywords: Biological systems, Estimation, Identification
Abstract: Elevated intracranial pressure (ICP) after traumatic brain injury can impede cerebral blood flow and contribute to secondary ischemic injury. However, direct ICP measurement is currently available only through invasive catheterization in critical care settings, while most tools being studied for noninvasive measurement permit only intermittent. This work examines the use of peripheral blood pressure waveform information to infer changes in ICP with a simple peripheral pressure sensor. Waveforms are reproduced by the direct fitting of a fluid circuit model for arterial and intracranial pressure behavior that was previously developed for use with an intraparenchymal catheter. A regression model is established between mean arterial blood pressure, differential ICP as inferred from the circuit model, and reference ICP measurements from invasive catheterization in eleven hospitalized subjects.

Keywords: LMIs, Control applications, Switched systems
Abstract: This paper addresses stability analysis of industrial high-precision nonlinear motion control systems using linear matrix inequalities (LMIs). Large-order state-space models required for accurate system representation can lead to computational challenges when solving the LMIs numerically. Using established tools from robust control theory, we present an approach that overcomes these limitations by using a low-order system approximation and treating the discrepancy between the approximation and the true system as an uncertainty. This uncertainty will be modeled in the frequency-domain. Stability for the low-order model is ensured through solving LMI conditions. Stability for the true system is guaranteed with an additional small-gain argument. The approach is demonstrated through a numerical example and an industrial case study on a lithography machine.

Keywords: LMIs, Optimization, Power generation
Abstract: This paper investigates a real-time optimization algorithm for autonomously calibrating the heliostats in a concentrated solar power plant to maximize power generation. The current state-of-the-art uses open-loop control with human-operators to provide feedback on the heliostats. This paper presents a method for tuning the RTO to ensure stable and fast convergence to the optimal alignment, which is robust to uncertainty in the shape of the sunspot. The exponential stability of the system is certified using a quadratic Lyapunov function and output feedback methods to couple the Lyapunov functions of the plant and the Real-Time Optimization algorithm (RTO). To validate stability, performance, and robustness, the closed-loop system is simulated using different power distributions that demonstrate the need for our RTO algorithm.

Keywords: Switched systems, LMIs, Stability of linear systems
Abstract: This paper proposes a convergent procedure for designing quadratically stabilizing switching rules for discrete-time switched systems. The technique relies on the computation of a Schur convex combination of a set of matrices by means of an LMI-based algorithm combined with simplicial partitioning of the unit simplex. Additionally, the utility of the method to improve other switching rules design strategies based on multiple Lyapunov functions is discussed. Numerical experiments demonstrate the effectiveness of the approach, providing comparisons with LMI-based techniques and periodic switching methods.

Keywords: Delay systems, Robust control, LMIs
Abstract: This paper addresses the robust stability of linear systems with unknown and distinct multi-input time-varying distributed delays. By utilizing multiple known nominal constant delays corresponding to each time-varying delay, reduction-type prediction-based control laws are proposed to robustly compensate for these time-varying distributed delays. Furthermore, we present a Lyapunov-based sufficient condition based on the linear matrix inequality (LMI) teachnology that demonstrates the exponential stability of the closed-loop system. In particular, the stability is ensured by sufficiently small variations between the constants and their corresponding time-varying distributed delays. Finally, a numerical example of a second-order linear system with two input channels is provided to verify the effectiveness of the proposed method.

Keywords: Constrained control, Optimal control, Linear systems
Abstract: The Linear Quadratic Regulator (LQR) is a classical problem in optimal control theory which deals with operating a linear dynamical system with optimized cost. In this work, we study the discrete-time LQR problem with sparsity constraints on the inputs. This problem has a combinatorial complexity. We develop a convex optimization-based approach to relax the problem into a semidefinite program which can be solved with polynomial complexity. We explore two cases for input sparsity: fixed temporal support and time-varying support. Moreover, we also solve the minimum-energy control problem with sparse inputs. Finally, using numerical simulations, we show that our algorithms give near-optimum performance with very good accuracy and time complexity.

Keywords: Stability of nonlinear systems, Observers for nonlinear systems, Uncertain systems
Abstract: This paper addresses optimal feedback stabilizing control for bounded Jacobian nonlinear discrete-time (DT) systems with nonlinear observations, affected by state and process noise. Instead of directly stabilizing the uncertain system, we propose stabilizing a higher-dimensional interval observer whose states enclose the true system states. Our nonlinear control approach introduces additional flexibility compared to linear methods, compensating for system nonlinearities and allowing potentially tighter closed-loop intervals. We also establish a separation principle, enabling independent design of observer and control gains, and derive tractable linear matrix inequalities, resulting in a stable closed-loop system.

Keywords: Power systems, Smart grid, Lyapunov methods
Abstract: Power grid stability assessment is a classic dynamic system problem which has become paramount with decarbonizing and fleeing dependence on fossil fuels. The relevant literature approaches studying power system stability using: (i) simplified, reduced order models that do not incorporate power flow constraints (and hence decouple generator transients from network topology), (ii) data-driven methods that are prone to measurement noise, (iii) inaccurate depictions of stochastic loads and renewables, (iv) specific metrics that are narrowly applied separately to various system states. This paper produces a fresh perspective---not necessarily superior---on stability assessment and uncertainty propagation by deriving generalizable stability metrics for assessing the impact of uncertain loads and renewables on network buses. The metrics use a purely model-driven nonlinear differential algebraic equation (NL-DAE) model. By generalizable, we mean that the metrics can be applied to various system states including frequencies, angles, and voltages. The proposed metrics are based on quantifying system stability via Lyapunov spectra of Exponents (LEs) of NL-DAEs. The metrics allow the identification of stable nodes, hence informing the operator on how uncertain load perturbations affect grid stability. We produce some case studies to demonstrate how this all works.

Keywords: Power systems, Decentralized control, Time-varying systems
Abstract: This paper presents a fast-acting, decentralized dynamic virtual inertial (VI) allocation method for frequency control in power grids with high penetration of inverter-connected resources under low and spatio-temporally varying inertia. The proposed decentralized method involves solving a local constrained convex optimization problem with implicit local frequency dynamics, enabling the use of projected gradient descent. The proposed controller stabilizes post-contingency frequency transients in milliseconds with less than 0.01% convergence error. Extensive simulations investigate the sensitivity of the cumulative and maximum frequency deviations, cumulative VI allocation, transient time and convergence error in frequency control to varying objective function weights on phase angle and frequency deviation, penalty on control effort, and gradient step sizes. The impact of this work is to propose decentralized control schemes as new mechanisms that act before or in alignment with primary and secondary control to safely regulate the frequency in future power grids dominated by inverter-connected resources.

Keywords: Smart grid, Power systems, Energy systems
Abstract: Electricity storage enhances the resilience of distribution systems against extreme weather-induced supply disruptions. Although recent insurance models with load shedding payouts have been proposed to mitigate outage impacts, it remains unclear whether such payouts (penalties) are necessary or if accurate outage probability estimates suffice. In this work, we model a profit-maximizing electricity retailer that invests in distributed storage under the risk of an interruption of bulk power supply during peak demand. We find that incorporating outage scenarios can reduce storage investments unless a load shedding penalty is imposed. We further demonstrate that proximity to demand, differences in net demand, and the relationship between wholesale and retail impact the locational marginal value of storage. This article highlights the importance of aligning incentives for the retailer. Without alignment, the retailer can expose consumers to ever greater damage costs as the risk of extreme weather increases.

Keywords: Control applications, Optimization, Modeling
Abstract: This study explores appropriate optimization algorithms to maximize power production and minimize temperature distribution in parallel collection lines in a parabolic trough solar collection process for energy generation. The objective is to minimize parasitic power losses from the oil circulation pump while maintaining discharge fluid in all loops at the set point temperature. Features of the objective function topography and how it changes during the day are revealed, and appropriate optimizers discussed. Results of four optimization algorithms are revealed: Generalized Reduced Gradient and Incremental Steepest Descent (both single trial solution, gradient based), Cyclic Heuristic (single trial solution, direct search) and Leapfrogging (multiplayer, direct search).

Keywords: Power systems, Identification
Abstract: In this paper, we quantify voltage and current phasor-based measurement requirements for the unique estimation of the electric grid topology and admittance parameters. Our approach is underpinned by the concept of a rigidity matrix that has been extensively studied in graph rigidity theory. Specifically, we show that the rank of the rigidity matrix is the same as that of a voltage coefficient matrix in a corresponding electric power system. Accordingly, we show that there is a minimum number of measurements required to uniquely estimate the admittance matrix and corresponding grid topology. By means of a numerical example on the IEEE 4-node radial network, we demonstrate that our approach is suitable for applications in electric power grids.

Keywords: Control applications, Energy systems, Emerging control applications
Abstract: As concerns grow about energy security and resilience, hybrid renewable energy systems (HRESs) are expected to play an important role in the future of energy. HRESs are comprised of multiple subsystems including electricity generation by renewable technologies, storage of energy or other products, and end uses including industrial loads. By coordinating the subsystem design and operation, significant potential exists to increase energy efficiency, even in situations with highly constrained transmission capacity, by taking advantage of complementarity among subsystems. This tutorial paper provides illustrative examples from three types of HRES: energy generation technologies, storage, and end use applications. All of these are situated in the broader, real-world context in which energy problems are defined and solved, with discussion about the impacts of this context on the technological solutions. The paper then describes a number of opportunities for control theory to advance the current HRES state-of-the art, including possible research questions to be addressed. Next, it describes a few of the many modeling and simulation tools that are available for HRES research. Finally, it concludes with discussion synthesizing the various elements of HRES with an aim to inspire future research.

Keywords: Energy systems, Hybrid systems, Hierarchical control
Abstract: To actively address the resource variability associated with variable renewable energy systems, in particular, wind power and solar power, recent entries into the interconnection queue (the queue of proposed power plants waiting to connect to the power grid) increasingly include the addition of battery storage or multiple generation sources. However, with such hybrid power plants coming online, specialized tools are needed to evaluate their performance and understand how they will contribute to grid stability and power supply-demand balancing. To address these needs, this tutorial article presents Hercules, a hybrid plant simulator developed at the National Renewable Energy Laboratory, as well as the Wind Hybrid Open Controller (WHOC), a library of baseline closed-loop controllers for such hybrid systems. In presenting Hercules and WHOC, we describe each at a high level to give readers an understanding of their purpose, basic functionality, and interactions, and then describe the inputs needed to run a simple simulation using the Hercules/WHOC ecosystem. We finish with an example simulation that demonstrates the tools in action and showcases tangible and intuitive results that highlight the potential benefits of using Hercules and WHOC for hybrid power plant evaluation and management.

Keywords: Autonomous systems, Learning, Game theory
Abstract: Deception is a key tactic for agents in adversarial environments, used to mislead opponents into adopting unaware strategies. In cyber-physical systems, for instance, deception can conceal attacks against critical infrastructure. This tutorial highlights the usefulness of deception for attacking and protecting systems against adversaries, but also as a tool to increase payoff in generic game-based and data-driven settings. It presents several state-of-the-art techniques for control-theoretic deception, including deception in defensive cyber-physical security, game-theoretic reinforcement learning, general multi-agent learning systems, Nash equilibrium seeking, and data-driven control. Although showcased in specific contexts, the underlying concepts and ideas that we study should be generalizable by researchers to settings beyond the scope of this tutorial.

Keywords: Autonomous systems, Aerospace, Optimization algorithms
Abstract: Aviation is undergoing a revolution and a paradigm change. New technologies are moving aviation towards on-demand transportation, often referred to as Advanced Air Mobility (AAM). To fully realize the promise of “anyone, anytime, anywhere” AAM transportation, autonomy must play a key role.

Our research team is focused on the intersection of new vehicle eVTOL configurations, popularly known as “air taxis”, and autonomous flight in complex urban environments. It has been widely recognized that dealing with contingencies, especially in a safe, scalable, and flexible way, is the most difficult challenge for autonomy. The tutorial session is intended to outline what we consider fundamental challenges and describe our current approaches. Moreover, we are working on establishing wide-ranging collaborations to address these fundamental autonomy challenges in a relevant environment with real-world assumptions and constraints. Hence, we would like to take this opportunity to discuss open challenge problems with this research community.

The main objective of this tutorial paper is to familiarize the audience with the inherent challenges of safe autonomous flight that is a required paradigm shift enabling economically feasible large-scale on-demand movement of cargo and people. To allow full exploitation of AAM, we consider eVTOL configurations, with their over-actuated multi-modal dynamics, operating in complex urban environments. There have been significant advances in robotics, artificial intelligence/machine learning, path-planning and control that advanced autonomous capabilities of aerial robots. However, the requirements of an integrated vehicle system capable of safe autonomous flight under both nominal and off-nominal conditions have not been addressed. The session is designed to outline compelling research challenges and to facilitate wider community engagement in advancing the science of autonomy for aviation.

Keywords: Distributed control, Optimization algorithms, Decentralized control
Abstract: In this work, we address the objective of protecting the states of a distributed dynamical system from eavesdropping adversaries. We prove that state-of-the-art distributed algorithms, which rely on communicating the agents' states, are vulnerable in that the final states can be perfectly estimated by any adversary including those with arbitrarily small eavesdropping success probability. While existing literature typically adds an extra layer of protection, such as encryption or differential privacy techniques, we demonstrate the emergence of a fundamental protection quotient in distributed systems when innovation signals are communicated instead of the agents' states.

Keywords: Optimization, Communication networks, Autonomous robots
Abstract: This paper tackles the challenge of resource allocation in large-scale datacenter networks (DCNs) using integer programming. The complexity of integer resource allocation problems (IRAP) lies in the non-convex nature of integer constraints, which makes achieving global optimality computationally challenging, especially in distributed and dynamic environments like DCNs. Traditional centralized methods struggle with scalability, efficiency, and adaptability to fluctuating workloads, which are common in data-intensive applications such as artificial intelligence (AI) and the Internet of Things (IoT). To address these challenges, we propose an asynchronous distributed resource allocation algorithm leveraging a gossip-based communication protocol. By transforming the resource allocation problem into an IRAP and using incremental cost functions, our method enables independent servers to make local decisions while converging to the global optimal solution. The algorithm is designed to overcome the difficulties of integer optimization in distributed networks, ensuring convergence through theoretical guarantees based on Markov processes. Simulations demonstrate the algorithm’s robustness, scalability, and effectiveness in various network scenarios.

Keywords: Optimization, Cooperative control, Optimization algorithms
Abstract: This paper addresses the problem of distributed optimization in an open network setting where new agents can join and existing ones can leave the network at specific times. Each agent has a local cost function, and the collective goal is to minimize the aggregate cost. As the number of agents is time-varying, the optimal solutions to this problem are time-dependent, making finding an exact solution generally challenging. In this paper, we will study this distributed optimization under a redundancy condition. Under this property, we will show that solving this time-varying optimization problem is essentially equivalent to solving a static optimization problem. Our main focus is to show that the classic distributed local gradient descent method in the client-server architecture will converge exactly to an optimal solution. In addition, we provide a formula to characterize the convergence properties of this method in the setting of open networks under the redundancy condition.

Keywords: Filtering, Estimation, LMIs
Abstract: This paper presents an analysis of noise propagation in distributed sensing systems and a novel modification of Hankel singular value-based optimization objectives for sensory system design. We analyze distributed sensing architectures in which sensors are spatially arrayed across a dynamical system and outputs are generated by projecting measurements onto spatial sensitivity patterns. We find a corresponding expression for the average output covariance which we use to modify Hankel singular value-based metrics for achievable system performance. This modification decouples the value of the adjusted metric from output scaling or linear dependence, which do not affect system performance. We then provide a semi-definite program that optimizes sensitivity patterns to maximize the adjusted minimum unstable Hankel singular value. Finally, we explore the result of applying this adjustment to the minimum unstable Hankel singular value using two simple examples.

Keywords: Distributed parameter systems, Sampled-data control, Stability of nonlinear systems
Abstract: This paper investigates the problem of boundary control for parabolic PDE-ODE cascade systems with distributed connections. A novel sampled-data boundary controller is proposed, and a new analytical framework is developed to establish the stability of the resulting closed-loop system. The framework includes the following three key features: 1) A state transformation is introduced for the concerned PDE-ODE system with non-strict feedback structure using a newly devised backstepping-forwarding technique. 2) A small gain theorem based method and a piecewise continuous Lyapunov approach are utilized to overcome the limitations of existing approaches which are applicable only to pure PDEs. 3) The input-to-state stability (ISS) approach is employed to derive an exponentially stable estimate of the closed-loop system's state in L^2 norm. Simulations validate the effectiveness of our approach.

Keywords: Predictive control for linear systems, Cooperative control, Distributed control
Abstract: A novel distributed Command Governor (CG) scheme, relying on cooperative distributed optimization, is presented for multi-agent networked systems subject to local and coupling constraints. Unlike existing non-cooperative distributed CG schemes, the proposed approach leverages a distributed optimization framework based on relaxation techniques and duality theory to enable agents to cooperatively contribute individually to the minimization of a global performance index. Furthermore, agents exchange only a vector that encodes the resource utilization of other agents, preserving privacy by avoiding the exchange of information about objective functions, constraints, or admissible commands. The effectiveness of the proposed approach is validated through an illustrative example involving vehicles operating in a 2D space that are subject to connectivity keeping coupling constraints.

Keywords: Traffic control, Predictive control for nonlinear systems, Transportation networks
Abstract: This paper designs traffic signal control policy for a network of signalized intersections without knowing the demand and parameters. Within a model predictive control (MPC) framework, the proposed control policy consists of an algorithm that estimates parameters and a one-step MPC that computes control inputs using estimated parameters. The algorithm switches between different terminal sets of the MPC to explore different regions of the state space, where different parameters are identifiable. The one-step MPC minimizes a cost that approximates the sum of squares of all the queue lengths within a constant and does not require demand information. We show that the algorithm can estimate parameters exactly in finite time, and the one-step MPC renders maximum throughput in terms of input-to-state practical stability. Simulations indicate better transient performance regarding queue lengths under our proposed policies than existing ones.

Keywords: Traffic control, Control applications, Feedback linearization
Abstract: This paper introduces a traffic management algorithm for heterogeneous highway corridors consisting of both human-driven vehicles (HVs) and autonomous vehicles (AVs). The traffic flow dynamics are modeled using the heterogeneous METANET model, with variable speed control employed to maintain desired vehicle densities and reduce congestion. To generate speed control commands, we developed a hybrid framework that combines feedback linearization (FL) and model predictive control (MPC), treating the traffic system as an over-actuated, constrained nonlinear system. The FL component linearizes the nonlinear dynamics, while the MPC component handles constraints by generating virtual control inputs that ensure control limits are respected. To address the over-actuated nature of the system, we introduce a novel constraint mapping algorithm within the MPC that links virtual control input constraints to the actual control commands. Additionally, we propose a real-time reference density generation method that accounts for both AVs and HVs to mitigate congestion. Numerical simulations were conducted for two scenarios: controlling only AVs and controlling both AVs and HVs. The results demonstrate that the proposed FL-MPC framework effectively reduces congestion, even when speed control is applied exclusively to AVs.

Keywords: Traffic control, Transportation networks, Networked control systems
Abstract: This paper analyzes the routing Nash equilibrium on a discrete-time traffic network consisting of a mix of regular/human-driven vehicles (RVs) and connected-autonomous vehicles (CAVs). We analyze the network in the context of a population game, where each population corresponds to a distinct origin-destination (OD) pair and vehicle type. We propose a novel evolutionary dynamic model governed by the Impartial Pairwise Comparison protocol, where our model satisfies a discrete-time notion of delta-dissipativity and preserves state feasibility for any time step. Drivers are assumed to adjust their strategies according to a static payoff mechanism, taken as the negative of the path travel time. We leverage these properties to derive sufficient conditions for stability of both homogenous traffic and mixed-traffic with uniform CAV headway. We then study a more realistic payoff model in which the CAV headway depends on the type of vehicle being followed. For this case, we propose a routing algorithm for the CAV payoff developed using feedback about the system and derive sufficient conditions under which a globally stabilizing controller exists. Furthermore, we formulate the control design as an optimization problem to improve the system convergence rate, and we support our findings with numerical simulations.

Keywords: Transportation networks, Power systems, Control applications
Abstract: Existing electric taxi services rely on charging infrastructure to maintain their daily operations. Unfortunately, severe power system disruptions, such as power rationing, can impose harsh constraints on e-taxi charging activities and significantly affect the service quality of e-taxis fleets. To address this issue, we design a framework for Energy-Aware e-taxi fleet coordination via dynamic charging Rate (EAR), to provide a satisfactory service quality while meeting energy conservation requirements. In this framework, an e-taxi fleet coordination algorithm is designed to provide sustainable service quality during pre-rationing, rationing, and post-rationing phases. The coordination problem across the three phases is modeled as separate multi-objective mixed-integer linear problems due to the distinct objectives of each phase. The proposed solution is evaluated with a comprehensive dataset for an existing e-taxi system and charging infrastructures including nearly 800 e-taxis. Our data-driven evaluation shows that EAR improves the ratio of served passengers by 37.0% during the power rationing phase compared with the state-of-the-art method, which does not consider disruptions in charging infrastructure when coordinating e-taxis.

Keywords: Autonomous vehicles, Traffic control, Optimization algorithms
Abstract: In this paper, we consider the problem of coordinating traffic light systems and connected automated vehicles (CAVs) in mixed-traffic intersections. We aim to develop an optimization-based control framework that leverages both the coordination capabilities of CAVs at higher penetration rates and intelligent traffic management using traffic lights at lower penetration rates. Since the resulting optimization problem is a multi-agent mixed-integer quadratic program, we propose a penalization-enhanced maximum block improvement algorithm to solve the problem in a distributed manner. The proposed algorithm, under certain mild conditions, yields a feasible and person-by-person optimal solution of the centralized problem. The performance of the control framework and the distributed algorithm is validated through simulations across various penetration rates and traffic volumes.

Keywords: Cooperative control, Autonomous systems, Traffic control
Abstract: We consider a mixed-traffic environment in transportation systems, where Connected and Automated Vehicles (CAVs) coexist with potentially non-cooperative Human-Driven Vehicles (HDVs). We develop a Cooperation Compliance Control (CCC) framework to incentivize HDVs to align their behavior with socially optimal objectives using a ``refundable toll'' scheme. This scheme achieves a desired compliance probability for all non-compliant HDVs through a feedback control mechanism combining global (social) with local (individual) components. We apply this scheme to the lane-changing problem, where a ``Social Planner'' provides references to the HDVs, measures their state errors, and induces cooperation compliance for safe lane-changing through a refundable toll approach. Simulation results are included to show the effectiveness of CCC in terms of improved compliance and lane-changing maneuver safety and efficiency when non-cooperative HDVs are present.

Keywords: Optimal control, Linear systems, Sampled-data control
Abstract: Data-enabled policy optimization (DeePO) is a newly proposed method to attack the open problem of direct adaptive Linear Quadratic Regulator (LQR). In this work, we extend the DeePO framework to the linear quadratic tracking (LQT) with offline data. By introducing a covariance parameterization of the LQT policy, we derive a direct data-driven formulation of the LQT problem. Then, we use gradient descent method to iteratively update the parameterized policy to find an optimal LQT policy. Moreover, by revealing the connection between DeePO and model-based policy optimization, we prove the linear convergence of the DeePO iteration. Finally, a numerical experiment is given to validate the convergence results. We hope our work paves the way to direct adaptive LQT with online closed-loop data.

Keywords: Optimal control, Optimization
Abstract: Differential Dynamic Programming (DDP) is one of the indirect methods for solving an optimal control problem. Several extensions to DDP have been proposed to add stagewise state and control constraints, which can mainly be classified as Augmented Lagrangian methods, active set methods, and barrier methods. In this paper, we use an interior point method, which is a type of barrier method, to incorporate arbitrary stagewise equality and inequality state and control constraints. We also provide explicit update formulas for all the involved variables. Finally, we apply this algorithm to example systems such as the inverted pendulum, a continuously stirred tank reactor, car parking, and obstacle avoidance.

Keywords: Optimization algorithms, H-infinity control, Robust control
Abstract: Earlier works have pointed out a connection between the design of optimization algorithms and H∞-control synthesis. It soon became clear that a key limiting factor in improving algorithmic convergence rates can be traced to the strict causality in the dynamics that make up the iterative structure of the algorithm. The starting point of the present work has been the realization that implicit algorithms may overcome the barrier imposed by strict causality, and thereby can achieve superior convergence rates. Implementation issues are considered and naturally lead to the use of proximal operators. The resulting implicit algorithms demonstrate superior convergence rates when compared to their explicit counterparts. The proposed framework is illustrated through the optimization of ill-conditioned functions and of functions that can be split into one part that is both strongly convex and Lipschitz smooth, and another whose proximal operator can be effectively evaluated.

Keywords: Optimization algorithms, Optimization
Abstract: We present a solution to a particular form of bilevel programming to solve a resource allocation problem for request-response systems. Our formulation is motivated by potential inefficiencies in allocating computational resources to incoming user requests in such systems. We optimize the trade-off between the financial cost of resources and the opportunity cost of unfulfilled user demand. Our bilevel formulation consists of an upper problem, which has a constraint value appearing in the lower problem. We derive an efficient, novel method for finding global solutions to the upper problem when user utilities are logarithmic. Our solution 1) determines the optimal total resource volume to allocate and 2) determines the optimal distribution of these resources across user requests.

Keywords: Optimization algorithms, Optimization
Abstract: We devise a novel quasi-Newton algorithm for solving unconstrained convex optimization problems. The proposed algorithm is built on our previous framework of the iteratively preconditioned gradient-descent (IPG) algorithm. IPG utilized Richardson iteration to update a preconditioner matrix that approximates the inverse of the Hessian matrix. In this work, we substitute the Richardson iteration with a successive over-relaxation (SOR) formulation. The convergence guarantee of the proposed algorithm and its theoretical improvement over vanilla IPG are presented. The algorithm is used in a mobile robot position estimation problem for numerical validation using a moving horizon estimation (MHE) formulation. Compared with IPG, the results demonstrate an improved performance of the proposed algorithm in terms of computational time and the number of iterations needed for convergence, matching our theoretical results.

Keywords: Optimization algorithms, Robust control, Lyapunov methods
Abstract: We show that the augmented primal-dual gradient algorithms can achieve global exponential convergence with partially strongly convex functions. In particular, the objective function only needs to be strongly convex in the subspace satisfying the equality constraint and can be generally convex elsewhere, provided the global Lipschitz condition for the gradient is satisfied. This condition implies that states outside the equality subspace will converge towards it exponentially fast. The analysis is then applied to distributed optimization, where the partially strong convexity can be relaxed to the restricted secant inequality condition, which is not necessarily convex. This work unifies global exponential convergence results for some existing centralized and distributed algorithms.

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: Space exploration embodies humanity's driver to transcend known boundaries and catalyses technological innovation, scientific advancements, and economic growth. As missions become increasingly complex, control theory emerges as a fundamental component, enhancing spacecraft navigation, operation, and adaptability in the dynamic space environment. This tutorial paper explores the pivotal role of control theory in advancing space exploration beyond the 2030s, emphasizing its contributions to autonomous decision-making, artificial intelligence, robust and resilient control, and the management of distributed systems. This paper outlines how advancements in control technologies will significantly enhance mission success and expand humanity's presence in the solar system from both fundamental research and commercial viewpoints, with contributions from academia, space agencies, and industry. By aligning with the strategic framework of the Global Exploration Roadmap, this tutorial paper presents a comprehensive framework for addressing current limitations in space exploration while devising future possibilities.

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: Artificial intelligence and machine learning (AI/ML) algorithms offer unprecedented opportunities in on-orbit servicing, space logistics and domain awareness. These opportunities include enhancing as well as enabling new Guidance, Navigation, and Control (GNC) tasks, autonomous collision avoidance at scale, and decision making for space operations and astronauts. At the same time, AI/ML algorithms face challenges for space applications that need to be tackled for their trusted deployment. These challenges are correlated and include data scarcity, a remote and harsh environment, risk adversity, and hard computational constraints. This presentation will provide an overview of the most recent research conducted at the Center for Aerospace Autonomy Research (CAESAR) at Stanford to unlock the aforementioned opportunities through a judicious incorporation of AI/ML algorithms in the autonomy stack for spacecraft Rendezvous and Proximity Operations (RPO). Among the key results in perception and navigation, it will be shown how a Vision Transformer can be trained online by an adaptive Kalman filter to increase the robustness of spacecraft pose estimation and tackle the so-called sim2real gap in RPO. It will also be shown how a spacecraft 3D structure can be rapidly abstracted from a single 2D image using an Auto-encoder network. Among the key results in guidance and control, it will shown how a novel Autonomous Rendezvous Transformer (ART) network can provide near-optimal warm-starts for a sequential convex program, leading to advantages both in terms of fuel optimality and computational efficiency in RPO in Earth as well as Cislunar orbits. Finally, preliminary results will be shown on how an open-source Large Language Model (LLM) pre-trained on internet-scale data can be adapted to on-orbit servicing applications and provide spatial reasoning and navigation around a target resident space object.

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: Future spacecraft and surface robotic missions require increasingly capable autonomy stacks for exploring challenging and unstructured domains and trajectory optimization will be a cornerstone of such autonomy stacks. However, the optimization solvers required remain too slow for use on resource constrained flight-grade computers. In this work, we present Transformer-based Powered Descent Guidance (T-PDG), a scalable algorithm for reducing the computational complexity of the direct optimization formulation of the spacecraft-powered descent guidance problem. T-PDG uses data from prior runs of trajectory optimization algorithms to train a transformer neural network, which accurately predicts the relationship between problem parameters and the globally optimal solution for the powered descent guidance problem. The solution is encoded as the set of tight constraints corresponding to the constrained minimum-cost trajectory and the optimal final landing time. By leveraging the attention mechanism of transformer neural networks, large sequences of time series data can be accurately predicted when given only the spacecraft state and landing site parameters. When applied to the real problem of Mars-powered descent guidance, T-PDG reduces the time for computing the 3 degrees of freedom fuel-optimal trajectory when compared to lossless convexification, improving solution times by up to an order of magnitude. A safe and optimal solution is guaranteed by including a feasibility check in T-PDG before returning the final trajectory.

Keywords: Chemical process control, Estimation, Control applications
Abstract: This paper addresses the challenge of estimating both states and parameters of post-combustion carbon capture plants (CCPs), with the goal of predicting the amount of captured CO2 using temperature measurements. We use a first-principle model of the CCP and employ simultaneous state and parameter estimation within a moving horizon estimation (MHE) framework. Sensitivity analysis and orthogonalization are used to select estimable states and parameters, enhancing es- timation accuracy and computational efficiency. Real industrial data is used to validate the proposed approach. Comparisons with other estimation methods highlight the effectiveness of the proposed approach. This work contributes practical insights into state and parameter selection, estimation methods for differential algebraic equation (DAE) systems, and data pre-processing in real-world settings.

Keywords: Modeling, Machine learning, Chemical process control
Abstract: This work proposes a modular learning framework that integrates individual process models into a global model for an entire process network. Three types of modules are discussed: black-box neural networks for processes with sufficient data, foundational modules using the reptile method for data-limited processes, and first-principles modules for processes with well-understood physicochemical phenomena. These modules are combined through an aggregation function to capture nonlinear dynamics of the entire process network. The proposed modular learning method, applied within a model predictive control framework, is demonstrated using a polymer production process, where it surpasses conventional neural networks in terms of accuracy, flexibility, and efficiency.

Keywords: Chemical process control, Predictive control for nonlinear systems, Machine learning
Abstract: This study introduces a dynamic data-driven model predictive control (MPC) framework for an ammonia synthesis packed-bed reactor (PBR). In response to the computational demands associated with high-fidelity modeling of the PBR, we propose a surrogate model based on long short-term memory (LSTM). This LSTM-based model adeptly captures the nonlinear dynamics of the process by leveraging information from offline computational fluid dynamics (CFD) simulations. While this surrogate model effectively mitigates the computational challenges associated with MPC, constraints emerge due to limitations in real-time measurements of maximum temperature and outlet ammonia concentration. To address these constraints, we present a feedforward neural network (FNN)-based state observer that utilizes temperature sensor values spatially distributed throughout the reactor's length to estimate maximum temperature and ammonia outlet concentration. The FNN-based state observer effectively estimates these key controlled outputs, facilitating MPC implementation and achieving optimal closed-loop performance in scenarios where real-time measurements are challenging to obtain.

Keywords: Chemical process control, Machine learning, Predictive control for nonlinear systems
Abstract: The complex structure of lignin presents significant challenges in understanding its reaction kinetics and optimizing its properties. Thus, multiscale kinetic Monte Carlo (kMC) models have been developed to provide detailed insights into the fractionation processes. The kMC calculates reaction rates between species, generates a rate-based probability distribution, and executes the most probable reactions at each time step. Despite its ability to handle nonlinear dynamics, the high computational demand associated with rate calculations makes real-time control challenging. Therefore, an artificial neural network (ANN) has been trained on the kMC input/output data to replace the repetitive rate calculation step of the kMC algorithm. This accelerated kMC model predicts reaction rates given the current status instead of calculating them, thereby significantly reducing the computational loads of the original kMC model. Integrated into a model predictive control (MPC) framework, the accelerated kMC enables real-time control of lignin properties, enhancing the efficiency and feasibility of the lignin fractionation process.

Keywords: Distributed parameter systems, Chemical process control, Predictive control for linear systems
Abstract: This paper presents the model predictive control of an axial tubular reactor with a recycle stream, where the intrinsic time delay imposed by the recycle stream—often overlooked in chemical engineering process control studies—is modeled as a transport PDE. This leads to a boundary-controlled system of coupled parabolic and hyperbolic PDEs under Danckwerts boundary conditions, specific for this reactor type. A discrete-time linear model predictive controller is designed to stabilize the system. Utilizing Cayley-Tustin time discretization along with the late lumping approach, the system's infinite-dimensional characteristics are preserved with no need for model reduction or spatial approximation. Numerical simulations demonstrate the controller's effectiveness in stabilizing an unstable system while satisfying input constraints.

Keywords: Control applications, Process Control, Grey-box modeling
Abstract: Cell culture processes for the production of therapeutic monoclonal antibodies (mAbs) often suffer from high batch-to-batch variability in the final concentration of the mAb (titer) in the bioreactor. In this study, we investigated the adoption of model predictive control (MPC) to reduce this variability, even under process disturbances. Both Linear MPC (LMPC) and Nonlinear MPC (NMPC) were tested in a Design of Experiments (DoE) involving 24 bioreactors at a 250 mL scale using human-in-the-loop (HITL) tracking MPC. The LMPC used a linear state-space model, while the NMPC employed a polynomial-NARX model to describe the process dynamics. Different estimators were applied for each MPC scheme. The effect of two different disturbances was evaluated. Experimental results showed that MPC significantly reduced process variation, with titer tracking error decreasing from a range of 18 percentage points without MPC to approximately 5 percentage points for both LMPC and NMPC case studies. These findings suggest that MPC can effectively control titer in bioprocessing, offering potential for increased efficiency and productivity.

Keywords: Lyapunov methods, Stability of nonlinear systems, Modeling
Abstract: A new class of oscillators is introduced as a second order nonlinear state-space equation with a clever structure that allows for analytical solutions. The periodic waveforms of these oscillators can be flexibly shaped by a function parameterizing the state-space equation. By restricting this function to a certain but still large class of functions, it is shown that the oscillators have a stable limit cycle described by a closed-form expression. In addition, the periodic waveforms are represented in terms of the parameterizing function via analytical expressions, which provide a powerful tool for shaping the waveforms arbitrarily by facilitating the parameter selection process. This process can be applied for design of new oscillators with complex waveforms or for modeling the existing, for instance, biological oscillators.

Keywords: Automotive control, Constrained control, Lyapunov methods
Abstract: Autonomous vehicles utilize low-level controllers to ensure vehicles stay within road boundaries while accurately tracking planned high-level reference trajectories. In this paper, we propose a control design that addresses both lateral reference tracking and lane-keeping safety objectives. This design exploits a Lyapunov Function-Based Reference Governor that handles control and safety constraints. We show that such a Reference Governor can be implemented without requiring iterative onboard optimization (i.e., it is optimization-free). Simulation results demonstrate that this approach can achieve accurate lateral reference tracking with safety guarantees. In comparison to an alternative solution that uses Control Barrier Functions and quadratic programming, the proposed method is able to generate larger constrained Domains of Attraction while requiring a shorter computation time of 0.12 pm 0.22~rm{ms}.

Keywords: Lyapunov methods, Stability of nonlinear systems, Adaptive systems
Abstract: Adaptive control techniques are ubiquitous methods for controlling dynamic systems, particularly because of their ability to improve system performance in the presence of uncertainties. However, a downside to these adaptive controllers is that particular learning rates are often required to ensure system performance requirements, creating high-frequency oscillations in the control input signal. These oscillations can potentially cause the system to become unstable or to have unacceptable performance. Thus, in this paper, we introduce a low-frequency learning adaptive control architecture for a discrete dynamical system with system uncertainties. In this framework, the update law is modified to include a filtered version of the updated parameter, allowing for high-frequency content to be removed while preserving system performance requirements. Lyapunov stability analysis is provided to guarantee asymptotic tracking error convergence of the closed-loop system. The results of a numerical simulation illustrate the reduction of high-frequencies in the system response.

Keywords: Lyapunov methods, Robust adaptive control, Neural networks
Abstract: Cable-driven exoskeletons have been used as training tools and to augment muscle effort during treadmill walking. However, technical challenges remain to achieve precise control of joints using such devices. Cable tension is controlled by an electric motor to rotate a joint; however, the mapping from the input current to applied joint torque is uncertain, nonlinear, and experiences a deadzone (DZ) due to cable slackness. Further, the cable-driven device needs to account for the nonlinear, non-parameterizable uncertainty in the target muscle-tendon complex. In this paper, a neural network (NN)-based controller is developed for a robotic cable-driven ankle-foot orthosis that targets lower-leg muscles while walking. First, a NN estimator and compensator are developed to mitigate the uncertain, nonlinear DZ that exists between the motor control input and the torque applied about the ankle. The motivation for the NNs is to inject a DZ-free controller in the closed-loop error system by generating a bounded estimation of an unknown DZ preinverse function. Then, robust control terms are designed along with an additional NN that exploits the desired kinematic trajectories to mitigate uncertainty in the dynamic model. A Lyapunov-based stability analysis is developed leveraging a corollary for non-smooth systems to establish an asymptotic kinematic tracking result.

Keywords: Nonlinear output feedback, Lyapunov methods, Stability of nonlinear systems
Abstract: In this paper, we introduce the notion of uniformly strongly dissipative dynamical systems and show that for a closed dynamical system (i.e., a system with the inputs and outputs set to zero) this notion implies fixed time stability. Specifically, we construct a stronger version of the dissipation inequality that implies system dissipativity and generalizes the notions of strict dissipativity and strong dissipativity while ensuring that the closed system is fixed time stable. The results are then used to derive extended Kalman-Yakubovich-Popov conditions for characterizing necessary and sufficient conditions for uniform strong dissipativity in terms of the system drift, input, and output functions using continuously differentiable storage functions and quadratic supply rates. Furthermore, using uniform strong dissipativity concepts we present several stability results for nonlinear feedback systems that guarantee finite time and fixed time stability. For specific supply rates, these results provide generalizations of the feedback passivity and nonexpansivity theorems that additionally guarantee finite time and fixed time stability.

Keywords: Stability of nonlinear systems, Lyapunov methods
Abstract: This paper focuses on stability tests for discrete-time nonlinear dynamical systems with a continuum of equilibria. Two notions that are of particular relevance to such systems are convergence and semistability. In this paper, we develop Lyapunov-based stability tests for decomposable discrete-time nonlinear systems using the notions of nontangency and restricted prolongations. These results do not make any assumptions on the sign definiteness of the Lyapunov function. Several examples are provided to illustrate how our results can be used for analyzing stability and convergence of systems having a continuum of equilibria.

Keywords: Decentralized control, Control over communications, Autonomous systems
Abstract: This paper proposes an approach to multi-agent collision avoidance that integrates Model Predictive Control (MPC) with intent communication. The work aims to address key challenges in decentralized multi-agent systems, where non-stationary dynamics and limited observability often lead to suboptimal strategies. The paper proved the C2I2 condition--- an at least local optimal path convergence with communication, infinite prediction horizon, and high prediction resolution, and a proper cost function---as a theoretical framework ensuring stability, path safety and priority.

Experimental results validate the theory, showing that the combination of MPC and communication reduces additional travel time by up to 41% and stuck cases by 30%, even under noisy and lossy communication conditions. The marginal difference between one-time communication and multi-round negotiation, along with the system's resilience under lossy and noisy conditions, highlights that the existence of communication itself is of primary importance, while the specifics of the communication method can be flexible, the result is available in the GitHub link.

Keywords: Optimization, Process Control, Predictive control for nonlinear systems
Abstract: This paper presents a constraint-oriented economic Nonlinear Model Predictive Control (eNMPC) framework designed to address multi-priority problems while ensuring asymptotic stability. One highlight is that the framework innovatively transforms different objectives into constraint forms and incorporates the lexicographic approach to solve multi-objective problems in a hierarchical manner. At the same time, asymptotic stability is ensured by embedding Lyapunov functions within the constraints. Another highlight is that this approach meets practical demands by enabling explicit priorities of objectives, thereby offering flexibility and stability in managing complex system requirements. The effectiveness of the proposed methodology is demonstrated through its application to a continuously stirred tank reactor (CSTR) system, showing the ability to manage conflicting priorities.

Keywords: Autonomous systems, Autonomous robots, Air traffic management
Abstract: In this paper, we develop a framework for the automatic taxiing of aircraft between hangar and take-off given a graph-based model of an airport. We implement a high-level path-planning algorithm that models taxiway intersections as nodes in an undirected graph, algorithmically constructs a directed graph according to the physical limitations of the aircraft, and finds the shortest valid taxi path through the directed graph using Dijkstra's algorithm. We then use this shortest path to construct a reference trajectory for the aircraft to follow that considers the turning capabilities of a given aircraft. Using high-order control barrier functions (HOCBFs), we construct safety conditions for multi-obstacle avoidance and safe reference tracking for simple 2D unicycle dynamics with acceleration control inputs. We then use these safety conditions to design an MPC-CBF framework that tracks the reference trajectory while adhering to the safety constraints. We compare the performance of our MPC-CBF controller with a PID-CBF control method via simulations.

Keywords: Energy systems, Process Control
Abstract: This paper presents dynamic simulation results of an MPC control strategy suitable for on-line control of a 100 kW SOFC-ICE hybrid stationary power generator in transient and steady-state operation.

Keywords: Chemical process control, Predictive control for nonlinear systems, Energy systems
Abstract: This paper investigates the regulation of a modular renewable energy-integrated hydrogen-ammonia production system comprising a water electrolyzer connected to an ammonia synthesis reactor with a recycle loop. This chemical production module is driven by an integrated renewable solar and wind energy generation system. Oscillatory behavior is observed in the chemical process, characterized by significantly faster dynamic responses compared to the renewable energy module. Under typical weather conditions, the renewable energy supply cannot fully meet the chemical system’s energy demands. Employing nonlinear model predictive control, this work examines optimal management strategies to maintain desired chemical manufacturing system operations based on the instantaneous availability of renewable energy. Specifically, the study identifies the minimum supplemental energy required from the electrical grid to achieve a target chemical operation.

Keywords: Distributed parameter systems, Differential-algebraic systems, Lyapunov methods
Abstract: This paper investigates the finite-dimensional observer-based boundary control for 1D linear parabolic-elliptic systems via the modal decomposition method. To address the potential multiple eigenvalues arising from the elliptic equation, we implement bilateral actuations (one Dirichlet and one Neumann) on the boundary of the parabolic equation with two point measurements. When the eigenvalues are simple, one boundary actuation and one point measurement are sufficient, but the second input and output may reduce the observer dimension. We present efficient LMI conditions for finding observer dimension, as well as controller and observer gains, ensuring the H^1 exponential stability with any desirable decay rate. We show that the LMIs are always feasible for large enough values of the observer dimension. Numerical examples demonstrate the efficiency of the method.

Keywords: Distributed parameter systems
Abstract: We provide two methods for computation of continuum backstepping kernels that arise in control of continua (ensembles) of linear hyperbolic PDEs and which can approximate backstepping kernels arising in control of a large-scale, PDE system counterpart (with computational complexity that does not grow with the number of state components of the large-scale system). In the first method, we provide explicit formulae for the solution to the continuum kernels PDEs, employing a (triple) power series representation of the continuum kernel and establishing its convergence properties. In the second method, we identify a class of systems for which the solution to the continuum (and hence, also an approximate solution to the respective large-scale) kernel equations can be constructed in closed form. We also present numerical examples to illustrate computational efficiency/accuracy of the approaches, as well as to validate the stabilization properties of the approximate control kernels, constructed based on the continuum.

Keywords: Agents-based systems, Distributed parameter systems, Cooperative control
Abstract: This paper introduces innovative finite-time and fixed-time state feedback stabilization techniques for systems modeled by inhomogeneous parabolic partial differential equations, with a focus on control strategies implemented either within the domain or at the boundaries. Diverging from previous studies, this research specifically examines the influence of an inhomogeneous source term in the PDE domain, particularly in scenarios where the source term is independent of the system state, output, and boundary conditions. The essential role of inhomogeneous parabolic PDE systems in various applications is examined. Initially, we address the problem of finite-time stabilization for inhomogeneous parabolic PDEs by first designing an in-domain controller for systems that are controllable within the domain, followed by the design of a boundary controller for systems controllable at the boundaries. Subsequently, we tackle the issue of fixed-time stabilization for inhomogeneous parabolic PDEs, where the convergence time is independent of the system states. This involves first designing an in-domain controller for systems controlled within the domain, and then developing a boundary controller for systems controlled at the boundaries. To evaluate the stability of the closed-loop system, the Lyapunov technique is employed for analysis. Finally, simulations are conducted to demonstrate the effectiveness of the proposed methodologies.

Keywords: Distributed parameter systems, Sampled-data control, Lyapunov methods
Abstract: This paper proposes an observer-based sampled-data boundary control approach for a class of reaction-diffusion PDEs with anti-collocated sensing and actuation. The PDE backstepping design is employed as the underlying control method. Using Lyapunov arguments, it is proven that continuous-time feedback boundary control, applied in a sample-and-hold manner, preserves closed-loop global exponential stability, provided that the sampling period is sufficiently small. The introduction of a state-independent dynamic reset variable in the Lyapunov function, to the best of our knowledge not previously employed in PDE control, enables the use of Lyapunov arguments to derive stability properties under sampled-data boundary control with PDE backstepping—previously only derived using small-gain arguments. The careful handling of this state-independent dynamic reset variable within the Lyapunov function absorbs the effect of input sampling, thereby preserving closed-loop global exponential stability. The results obtained provide stability estimates for the spatial L^2 norms of the plant and observer states. Furthermore, robustness to perturbations in the sampling schedule is guaranteed by establishing a suitable maximum upper diameter for the sampling interval. A numerical simulation is provided to illustrate the theoretical results.

Keywords: Distributed parameter systems, Adaptive systems
Abstract: This paper deals with modifications to the standard adaptive law for finite time convergence of parameters in on-line estimation of spatial fields. The motion of a single sensor providing observations of the spatial field is a necessary and sufficient condition for the parameter convergence as its motion ensures persistence of excitation. The additional condition of positivity of filtered values of the trajectory-dependent regressor matrix which ensures finite time convergence, is shown to be reduced to a trajectory design of a mobile sensor.

Keywords: Distributed parameter systems, Fault accomodation, Process Control
Abstract: This work focuses on the problem of mitigating control actuator faults in spatially distributed processes modeled by highly dissipative Partial Differential Equations (PDEs) and controlled using periodically sampled and delayed measurements. Initially, a stabilizing model-based state feedback controller that takes account of measurement sampling and measurement delays is designed using a suitable low-order approximation of the infinite-dimensional system. A rigorous characterization of the closed-loop stability properties linking the faults and the controller parameters is obtained. This characterization serves as a basis for the development of fault mitigation strategies that maintain closed-loop stability in the presence of faults. In addition to stability considerations, a rigorous characterization of the closed-loop performance properties linking the extended H_2-norm of the performance output to the faults and controller parameters is also obtained. This characterization provides the foundation for devising fault mitigation strategies aimed at maintaining closed-loop stability despite faults and minimizing performance decline after faults occur.

Keywords: Adaptive control, Machine learning, Manufacturing systems
Abstract: This work proposes the implementation of a machine-learning endpoint detector system as a feedback controller for the process time of an atomic layer etching process of a silicon wafer. The endpoint feedback control system is based on a Transformer soft-sensor model that uses real-time pressure readings from multiple points on the wafer surface to predict whether the wafer has been fully processed. These pressure readings are continuously read throughout the entire process, making this a real-time feedback control system with an input that has a variable sequence length. To evaluate the endpoint control system, various Transformer models that were trained and tested on unique sets of process data were considered. These various sets of process data were created from multiscale computational fluid dynamics (CFD) process simulations that are intended to represent industrial process data. By comparing the endpoint control system's predicted end time and the actual optimal end time, the effectiveness of the Transformer model can be quantified for a variety of situations specified using industrial feedback. The results indicate that the endpoint control system's performance decreases drastically if the Transformer model is not trained on a range of process data that contains the disturbances it is expected to encounter. Because industrial data typically encompasses years of data, common process disturbances are already included; thus, the proposed endpoint detector is a viable feedback control system that can be implemented in an industrial manufacturing environment.

Keywords: Adaptive control, Optimization, Uncertain systems
Abstract: We propose a dual-mode extremum seeking control design technique that achieves real-time optimization of an unknown measured cost function in a prescribed time. The controller is shown to achieve prescribed-time semi-global practical stability of the optimal equilibrium for the state variables and the input variable for a class of nonlinear dynamical control systems with unknown dynamics. The design technique proposes a timescale transformation that enables the use of dither signals with increasing frequencies. The proposed timescale transformation is designed to avoid the singularity occurring at the prescribed time. A simulation study is performed to illustrate the effectiveness of the proposed technique.

Keywords: Adaptive control, Reinforcement learning, Stochastic systems
Abstract: Autonomous systems must have the ability to quickly adapt to various situations. However, adaptation methods often require strong assumptions about system structures, environmental homogeneity, and multiple rollouts. In this work, we integrate multi-armed bandit and model-based RL to design a fast adaptation algorithm on a single trajectory. Our approach achieves sublinear regret of O(sqrt{T}), and the performance guarantee does not require homogeneity of the environment. This regret bound is achieved using a novel prediction error metric that is minimized in the ground-truth MDP. To the best of our knowledge, all existing results with provable guarantees depend on the Bregman divergence between the optimal policies among the MDPs. We show by simulation that our algorithm performs well in puzzle navigation and quadcopter path-tracking.

Keywords: Adaptive control, Neural networks, Learning
Abstract: Control Barrier Functions (CBFs) enforce safety in dynamical systems by ensuring trajectories stay within prescribed safe sets. However, traditional CBFs often overlook realistic input constraints, potentially limiting their practical effectiveness. To address this research gap, we introduce the Pareto Control Barrier Function (PCBF) algorithm, which employs a Pareto multi-task learning framework to simultaneously balance the competing objectives of maintaining safety and maximizing the volume of the safe set under input limitations. The PCBF algorithm is computationally efficient and scalable to high-dimensional systems. We demonstrate its effectiveness through comparisons with Hamilton-Jacobi reachability on an inverted pendulum and extensive simulations on a 12-dimensional quadrotor system, where PCBF consistently outperforms existing methods by yielding larger safe sets while ensuring robust safety under input constraints. Codes are available at https://github.com/XiaoyangCao1113/Pareto_CBF.

Keywords: Adaptive control, Control applications, Aerospace
Abstract: Autopilots for quadcopters are typically designed with an inner-outer loop architecture consisting of cascaded P and PID controllers. While straightforward, this design requires tuning numerous gains through simulation and flight tests, and may not be sufficiently robust to disturbances. To alleviate tuning requirements and to provide greater robustness, the present paper proposes an autopilot based on predictive cost adaptive control (PCAC). As an indirect adaptive control extension of model predictive control (MPC), PCAC uses recursive least squares (RLS) with variable-rate forgetting (VRF) for online system identification. The present paper compares PCAC with fixed-gain PID control in both simulation and flight tests with and without disturbances, and the performance improvement is examined.

Keywords: Adaptive control, Nonlinear output feedback, Stability of nonlinear systems
Abstract: This paper presents a numerical investigation of the ability of predictive cost adaptive control (PCAC) to stabilize self-excited systems modeled by discrete-time Lur'e systems. The closed-loop Lur'e system is comprised of the positive feedback interconnection of the Lur'e system and the PCAC controller. This work presents a numerical investigation of the circle and Tsypkin absolute stability criteria to evaluate the stability of the closed-loop Lur'e system at each step. An exogenous input is used to increase persistency and thus enhance the ability of the closed-loop system to satisfy the absolute stability criteria. A numerical example illustrates the effect of exogenous persistency on the stability of the closed-loop Lur’e system. These numerical results show the potential value of absolute stability criteria for assessing the ability of adaptive model predictive control to stabilize a class of nonlinear systems.

Keywords: Observers for Linear systems, Neural networks, LMIs
Abstract: This paper presents a novel approach for state and unknown input (UI) observer in single-input single-output (SISO) linear time-invariant systems (LTI). By incorporating an Orthogonal Neural Network (ONN) to approximate the unknown input, we eliminate the need for traditional training or offline learning. Our method extends the system state to include the ONN weights, creating a new LTI system. Leveraging the structure of orthogonal activation functions, we derive a suitable algorithm for the extended state dynamics. The observer gain is computed using Linear Matrix Inequalities (LMI), ensuring stability and attenuating the approximation error. Unlike existing methods, our approach requires no assumptions about the unknown input, such as matching conditions or boundedness. A numerical example demonstrates the effectiveness of the proposed scheme.

Keywords: Observers for nonlinear systems, Algebraic/geometric methods
Abstract: A symmetry-preserving, reduced-order state observer is presented for the unmeasured part of a system's state, where the nonlinear system dynamics exhibit symmetry under the action of a Lie group. Leveraging this symmetry with a moving frame, the observer dynamics are constructed such that they are invariant under the Lie group's action. Sufficient conditions for the observer to be asymptotically stable are developed by studying the stability of an invariant error system. As an illustrative example, the observer is applied to the problem of rigid-body velocity estimation, which demonstrates how exploiting the symmetry of the system can simplify the stabilization of the estimation error dynamics.

Keywords: Observers for nonlinear systems, Estimation, LMIs
Abstract: This paper suggests a new discrete-time nonlinear observer design based on an output dynamic extension technique. This method aims to filter the output measurements by minimizing the impact of measurement noise, thereby, enhancing the accuracy and reliability of state estimates. In order to ensure the Input-to-State Stability (ISS) property of the estimation error, a novel LMI condition is proposed. An application on an agricultural quadcopter model operating in an Adaptive Vertical Farm showcases the effectiveness of the proposed estimation approach.

Keywords: Observers for nonlinear systems, Indirect adaptive control, Learning
Abstract: This study proposes the design of learning-based nonlinear discrete-time Luenberger observers for the global reconstruction of discrete-time multi-tone sinusoidal signals with unknown frequencies while avoiding the use of adaptive techniques. The unknown parameters are estimated directly using an explicit nonlinear mapping, which achieves exponential convergence to the true unknown parameters. The proposed observer is applied in the design of a feedforward controller that solves the output regulation problem.

Keywords: Observers for nonlinear systems, Stochastic systems, Computational methods
Abstract: This paper proposes a method to synthesize robust and asymptotically stable-in-probability observers for stochastic discrete-time nonlinear systems, more specifically polynomial dynamical systems. Beyond synthesizing, the presented framework allows us to attain the characteristics of the observer, such as its convergence and robustness. While our analysis focuses on polynomial dynamical systems, it is important to note that most real-world systems can be effectively approximated by polynomial dynamics within a suitable region. This makes our approach widely applicable and not overly restrictive in practice. Numerical examples illustrate the effectiveness of the proposed observer.

Keywords: Robust control, Predictive control for linear systems, Computational methods
Abstract: This paper introduces pycvxset, a new Python package to manipulate and visualize convex sets. We support polytopes and ellipsoids, and provide user-friendly methods to perform a variety of set operations. For polytopes, pycvxset supports the standard halfspace/vertex representation as well as the constrained zonotope representation. The main advantage of constrained zonotope representations over standard halfspace/vertex representations is that constrained zonotopes admit closed-form expressions for several set operations. pycvxset uses CVXPY to solve various convex programs arising in set operations, and uses pycddlib to perform vertex-halfspace enumeration. We demonstrate the use of pycvxset in analyzing and controlling dynamical systems in Python. pycvxset is available at https://github.com/merlresearch/pycvxset under the AGPL-3.0-or-later license, along with documentation and examples.

Keywords: Robotics, Autonomous systems, Model/Controller reduction
Abstract: This paper presents a method for exactly representing the free space around static, convex, polytopic obstacles as a collection of disjoint, convex polytopes. Connections between hyperplane arrangements, shallow ReLU neural networks, and hybrid zonotopes are used to efficiently partition the free space into convex cells. A greedy algorithm is presented for merging neighboring cells, while preserving convexity, to reduce the number of cells needed to represent the obstacle-free space. Compared to two existing methods, the proposed approach is shown to provide a practical balance between minimizing the number of cells representing the obstacle-free space and achieving computational scalability with respect to the number of obstacles in two-dimensional examples.

Keywords: Uncertain systems, Identification, Optimization
Abstract: Ensuring the safety of real systems necessitates the identification of non-deterministic models that capture all system behaviors. This process is often referred to as reachset-conformant identification. The state of the art, however, relies on batch processing of data, resulting in a) large linear programs to be solved when dealing with large datasets and b) in overly conservative solutions when identifying time-variant systems. We address these problems by proposing recursive reachset-conformant identification methods, which iteratively refine the identification results. Amongst others, we present a novel algorithm to reduce the number of linear constraints in general optimization problems by underapproximating the feasible set of solutions. Additionally, we enable the models to adapt to changing system dynamics by incorporating forgetting factors in our methods. The proposed methods are evaluated by numerical experiments that show significant improvements in computational efficiency with only slight increases in conservatism.

Keywords: Computational methods, Linear systems, Uncertain systems
Abstract: This paper introduces memZono, a new class within the zonoLAB MATLAB toolbox which exploits the memory encoded within zonotope-based set representations to enable dimensionally-aware and dependency-preserving set operations. The factor-based construction of zonotope-based sets inherently encodes the origin of set contributions; however, without careful usage this information becomes scrambled under set operations. To formally track and maintain the set contributions, novel set operations are introduced that align the dimensions, factors, and constraints appropriately. The memZono class aims to make these advanced memory-preserving capabilities easily accessible to users without the need to handle the bookkeeping themselves. A reachability example is used as a motivation and tutorial for memZono functionality and usage. Additional applications to simultaneous localization and mapping (SLAM) and functional creation further demonstrate the wide utility, flexibility, and power of working with the zonotope memory enabled by memZono.

Keywords: Aerospace, Predictive control for linear systems, Energy systems
Abstract: Uncrewed aerial systems have tightly coupled energy and motion dynamics which must be accounted for by onboard planning algorithms. This work proposes a strategy for coupled motion and energy planning using model predictive control (MPC). A reduced-order linear time-invariant model of coupled energy and motion dynamics is presented. Constrained zonotopes are used to represent state and input constraints, and hybrid zonotopes are used to represent non-convex con- straints tied to a map of the environment. The structures of these constraint representations are exploited within a mixed-integer quadratic program solver tailored to MPC motion planning problems. Results apply the proposed methodology to coupled motion and energy utilization planning problems for 1) a hybrid-electric vehicle that must restrict engine usage when flying over regions with noise restrictions, and 2) an electric package delivery drone that must track waysets with both position and battery state of charge requirements. By leveraging the structure-exploiting solver, the proposed mixed-integer MPC formulations can be implemented in real time.

Keywords: Stochastic systems, Autonomous systems
Abstract: We study the safety verification problem for discrete-time stochastic systems. We propose an approach for safety verification termed set-erosion strategy that verifies the safety of a stochastic system on a safe set through the safety of its associated deterministic system on an eroded subset. The amount of erosion is captured by the probabilistic bound on the distance between stochastic trajectories and their associated deterministic counterpart. Building on recent development of stochastic analysis, we establish a sharp probabilistic bound on this distance. Combining this bound with the set-erosion strategy, we establish a general framework for the safety verification of stochastic systems. Our method is flexible and can work effectively with any deterministic safety verification techniques. We exemplify our method by incorporating barrier functions designed for deterministic safety verification, obtaining barrier certificates much tighter than existing results. Numerical experiments are conducted to demonstrate the efficacy and superiority of our method.

Keywords: MEMs and Nano systems, Mechatronics, Optimal control
Abstract: This paper presents the design process and compares three damping controllers for an uncertain microelectromechanical system (MEMS) force sensor to mitigate its resonant mode. The damping methods include resonant controller (RC), integral resonant controller (IRC), and robust resonant controller (RRC). A frequency domain-based convex optimization is proposed to find the parameters of the resonant controller to obtain a desired closed-loop response. In addition to the conventional IRC, we formulate a second-order robust resonant controller. To validate the effectiveness of the proposed controllers, we conduct comparative experiments under uncertainty where system dynamics are perturbed, resulting in a decrease in the system's DC-gain and an increase in the resonance frequency over a relatively wide bandwidth. The findings reveal that the RC exhibits superior damping characteristics and the shortest settling time. In contrast, the IRC shows minimal overshoot in step response (30%), and the RRC demonstrates enhanced robustness under uncertain conditions.

Keywords: Constrained control, Stability of nonlinear systems, Data driven control
Abstract: Invariant sets are crucial for ensuring safety and constraint adherence for dynamical systems. However, determining an invariant set for a dynamical system can be difficult. In the case of unmodeled dynamical systems, methods often rely on stochastic guarantees or assumptions that the model is linear. In contrast, we develop a deterministic sampling method to synthesize true invariant sets for unmodeled, Lipschitz continuous, discrete-time, potentially nonlinear dynamical systems from successive sampled state pairs {x,x^+}. The Lipschitz continuity of the system is leveraged to characterize the evolution of the system around the sampled points and identify regions of the state space where the successor set maps. A tree data structure is used for ease of computation to divide the space and identify each region. Our sampling scheme requires initialization with a small invariant set. We use the assumption that the region of interest's interior contains an exponentially stable equilibrium point to create a small candidate invariant set about an estimated equilibrium and provide an iterative sampling scheme to confirm the invariance of the initial set. Our sampling method then iteratively expands the initial invariant set until the area of the invariant set can no longer increase.

Keywords: Computational methods, Sampled-data control, Linear systems
Abstract: In our recent article [Chen et al., SIAM J Control Optim], we introduced a system-theoretic approach to a class of discrete-time multi-linear time-invariant (MLTI) dynamical systems, where the states, inputs, and outputs are all tensors. The purpose of this article is to delve into the role of data for MLTI systems. We perform data-driven analysis, concerning system identification, stability, controllability, and stabilizability, of MLTI systems. In particular, we exploit advanced tensor decomposition techniques encompassing CANDECOMP/PARAFAC decomposition and tensor train decomposition to establish effective criteria for determining the data informativity for the aforementioned system properties. We further demonstrate our framework with numerical examples.

Keywords: Linear systems, Behavioural systems
Abstract: In this paper, a data-driven direct input synthesis control (DD-DISC) approach is proposed for gls{LTI} gls{SISO} system. Recently, data-driven control (DDC) has attracted more and more attention to address limitations and challenges in model-based control for practical implementations, particularly, those related to modeling. Unlike the existing DDC techniques that are focused on constructing/representing a feedback controller using data, we propose to directly synthesize the control input for tracking a finite-previewed desired trajectory using past input-output data. Such a direct input synthesis approach avoids the complexity and constraints in designing a general-purpose controller, particularly for non-minimal phase systems. Specifically, the proposed approach addresses the issue and limitation of the recently-developed data-driven predictive control technique that depends on the observability of the system. The desired trajectory is partitioned and constructed as the extended desired trajectory, and the non-minimal phase effect of the system is considered. The robustness performance under the effect of the output noise/disturbance is analyzed, and the proposed DD-DISC method is illustrated through a simulation of non-periodic tracking on a piezoelectric actuator.

Keywords: Linear systems, Optimization algorithms, Optimization
Abstract: Minimum-gain pole placement is a classical problem that aims to find a static state feedback matrix with the minimum norm that places the closed-loop poles at desired locations. In this paper, we present the direct data-driven formulation of this problem without identifying the system model. We derive and discuss the conditions for pole placement using data matrices, and propose a projection-based gradient descent algorithm to solve the problem. We also consider sparsity constraints on the feedback matrix and obtain approximately sparse solutions. Our simulations show that the proposed direct method is more accurate than the model-based approach in placing the poles as well as in obtaining a feedback matrix with lower norm.

Keywords: Neural networks, Fuzzy systems, Constrained control
Abstract: The truck towing trailer(s) system is crucial in the postal and shipping industries. An N-trailer wheeled mobile robot is a nonlinear, under-actuated system. Assuming pure rolling of the robot's wheels, the system is subjected to nonholonomic constraints, self-collision risks, and actuator saturation limitations. To address these challenges, we designed a nonlinear model predictive controller (or NMPC) using a trapezoidal integrator. A fuzzy controller was also developed, which sets the maximum speed for each motor and incorporates a method to prevent self-collision. A data-driven controller was created based on a dataset generated from multiple NMPC runs. Finally, using Simulink, we compared the three controllers in terms of control performance, self-collision avoidance, actuator saturation, and computational load. The simulation results showed that the neural network (NN) struggled with constraint adherence, the NMPC had a high computational burden, and the fuzzy controller exhibited minor steady-state errors.

Keywords: Optimal control, Large-scale systems, Neural networks
Abstract: Convolutional neural networks (CNNs) have achieved remarkable success in representing and simulating complex spatio-temporal dynamic systems within the burgeoning field of scientific machine learning. However, optimal control of CNNs poses a formidable challenge, because the ultra-high dimensionality and strong nonlinearity inherent in CNNs render them resistant to traditional gradient-based optimal control techniques. To tackle the challenge, we propose an optimal inferential control framework for CNNs that represent a complex spatio-temporal system, which sequentially infers the best control decisions based on the specified control objectives. This reformulation opens up the utilization of sequential Monte Carlo sampling, which is efficient in searching through high-dimensional spaces for nonlinear inference. We specifically leverage ensemble Kalman smoothing, a sequential Monte Carlo algorithm, to take advantage of its computational efficiency for nonlinear high-dimensional systems. Further, to harness graphics processing units (GPUs) to accelerate the computation, we develop a new sequential ensemble Kalman smoother based on matrix variate distributions. The smoother is capable of directly handling matrix-based inputs and outputs of CNNs without vectorization to fit with the parallelized computing architecture of GPUs. Numerical experiments show that the proposed approach is effective in controlling spatio-temporal systems with high-dimensional state and control spaces.

Keywords: Neural networks, Machine learning, Nonlinear systems identification
Abstract: Extracting dynamic models from data is of enormous importance in understanding the properties of unknown systems. In this work, we employ Lipschitz neural networks, a class of neural networks with a prescribed upper bound on their Lipschitz constant, to address the problem of data-efficient nonlinear system identification. Under the (fairly weak) assumption that the unknown system is Lipschitz continuous, we propose a method to estimate the approximation error bound of the trained network and the bound on the difference between the simulated trajectories by the trained models and the true system. Empirical results show that our method outperforms classic fully connected neural networks and Lipschitz regularized networks through simulation studies on three dynamical systems, and the advantage of our method is more noticeable when less data is used for training.

Keywords: Neural networks, Machine learning, Optimal control
Abstract: In this paper, we propose a method to solve nonlinear optimal control problems (OCPs) with constrained control input in real-time using neural networks (NNs). We introduce what we have termed co-state Neural Network (CoNN) that learns the mapping from any given state value to its corresponding optimal co-state trajectory based on the Pontryagin’s Minimum (Maximum) Principle (PMP). In essence, the CoNN parameterizes the Two-Point Boundary Value Problem (TPBVP) that results from the PMP for various initial states. The CoNN is trained using data generated from numerical solutions of TPBVPs for unconstrained OCPs to learn the mapping from a state to its corresponding optimal co-state trajectory. For better generalizability, the CoNN is also trained to respect the first-order optimality conditions (system dynamics). The control input constraints are satisfied by solving a quadratic program (QP) given the predicted optimal co-states. We demonstrate the effectiveness of our CoNN-based controller in a feedback scheme for numerical examples with both unconstrained and constrained control input. We also verify that the controller can handle unknown disturbances effectively.

Keywords: Neural networks, Machine learning, Identification
Abstract: This paper proposes a minimalist and interpretable recurrent neural-network structure using the echo-state network (ESN) paradigm for time-series prediction. While the traditional ESNs perform well for dynamical systems prediction, it needs a large dynamic reservoir with increased computational complexity. It also lacks interpretability to discern contributions from different input combinations to the output. Here, a systematic reservoir architecture is developed using smaller parallel reservoirs driven by different input combinations, known as features, and then they are nonlinearly combined to produce the output. The resultant feature-based ESN (Feat-ESN) outperforms the traditional single-reservoir ESN with less reservoir nodes. The predictive capability of the proposed architecture is demonstrated on three systems: two synthetic datasets from chaotic dynamical systems and a set of real-time traffic data.

Keywords: Formal verification/synthesis, Hybrid systems, Neural networks
Abstract: Functional decomposition is a powerful tool for systems analysis because it can reduce a function of arbitrary input dimensions to the sum and superposition of functions of a single variable, thereby mitigating (or potentially avoiding) the exponential scaling often associated with analyses over high-dimensional spaces. This paper details the construction of functional decompositions for functions of arbitrary input and output dimensions, as well as providing algorithms for removing redundancies and excessive unary decompositions. To demonstrate the applications of functional decomposition, we construct a hybrid zonotope set that over-approximates the input-output graph of a long short-term memory neural network, and use functional decomposition to describe a new technique for the set-based reachability analysis of discrete hybrid automata.

Keywords: Adaptive control, Neural networks, Sampled-data control
Abstract: This paper presents a safety-critical control framework for an ego agent moving among other agents. The approach infers the dynamics of the other agents and incorporates the inferred quantities into control barrier function (CBF)-based controllers for the ego agent. The inference method combines offline and online learning with radial basis function neural networks (RBFNNs). The RBFNNs are initially trained offline, and their weights are updated online with new observations. Additionally, we employ adaptive conformal prediction to quantify the estimation error of the RBFNNs for the other agents' dynamics, generating prediction sets to cover the true value with a desired confidence level. Finally, we formulate a CBF-based controller for the ego agent to guarantee safety with the desired confidence level by accounting for the prediction sets of other agents' dynamics in the sampled-data CBF conditions. Simulation results are provided to demonstrate the effectiveness of the proposed method.

Keywords: Model Validation, Autonomous systems, Neural networks
Abstract: This paper presents a method to efficiently reduce conservativeness in reachable set over approximations (RSOAs) to verify safety for neural feedback loops (NFLs), i.e., systems that have neural networks in their control pipelines. While generating RSOAs is a tractable alternative to calculating exact reachable sets, RSOAs can be overly conservative, especially when generated over long time horizons or for highly nonlinear NN control policies. Refinement strategies such as partitioning or symbolic propagation are typically used to limit the conservativeness of RSOAs, but these approaches come with a high computational cost and often can only be used to verify safety for simple reachability problems. This paper presents Constraint-Aware Refinement for Verification (CARV): an efficient refinement strategy that reduces the conservativeness of RSOAs by explicitly using the safety constraints on the NFL. Unlike existing approaches that seek to to refine RSOAs over the entire time horizon, CARV limits the computational cost of refinement by refining RSOAs only where necessary to verify safety. We demonstrate that CARV can verify the safety of an NFL where other approaches either fail or take more than 60× longer and 40× the memory.

Keywords: Constrained control, Machine learning, Supervisory control
Abstract: An approach to safe and fast online learning of constraints for a continuous-time linear system subject to linear inequality constraints is developed, assuming that the number of constraints is known and measurements of the constraint signals are available. During the identification phase, a constant reference command input is applied for the duration of an epoch and constraint measurements are collected. Based on these measurements, the set of feasible constraint parameters is refined using set-membership learning techniques. The reference command value is selected so that it minimizes the worst-case uncertainty in the parameters after one epoch while safety is ensured through the use of appropriately defined safe sets. The characterization of safe sets is shown to reduce to a finite set of linear inequality constraints. A numerical case study is reported for the proposed algorithm.

Keywords: Stability of nonlinear systems, Lyapunov methods, Constrained control
Abstract: Analyzing and certifying the stability and attractivity of nonlinear systems is a topic of ongoing research interest that has been extensively investigated by control theorists and engineers for many years. However, accurately estimating domains of attraction for nonlinear systems remains a challenging task, where existing estimation methods tend to be conservative or limited to low-dimensional systems. In this work, we propose an iterative approach to accurately underapproximate safe (state-constrained) domains of attraction for general discrete-time autonomous nonlinear systems. Our approach relies on implicit representations of safe backward reachable sets of initial safe regions of attraction, where such initial regions can {be} easily constructed using, e.g., quadratic Lyapunov functions. The iterations of our approach are monotonic (in the sense of set inclusion), {converging to the safe domain of attraction}. Each iteration results in a safe region of attraction, represented as a sublevel set, that underapproximates the safe domain of attraction. { The sublevel set representations of the resulting regions of attraction can be efficiently utilized in verifying the inclusion of given points of interest in the safe domain of attraction}. We illustrate our approach through two numerical examples, involving two- and four-dimensional nonlinear systems.

Keywords: Data driven control
Abstract: This work is concerned with developing a data-driven approach for learning control barrier certificates (CBCs) and associated safety controllers for discrete-time nonlinear polynomial systems with unknown mathematical models, guaranteeing system safety over an infinite time horizon. The proposed approach leverages measured data acquired through an input-output observation, referred to as a single trajectory, collected over a specified time horizon. By fulfilling a certain rank condition, which ensures the unknown system is persistently excited by the collected data, we design a CBC and its corresponding safety controller directly from the finite-length observed data, without explicitly identifying the unknown dynamical system. This is achieved through proposing a data-based sum-of-squares optimization (SOS) program to systematically design CBCs and their safety controllers. We validate our data-driven approach over two physical case studies including a jet engine and a Lorenz system, demonstrating the efficacy of our proposed method.

Keywords: Linear parameter-varying systems, Data driven control
Abstract: This paper presents a data-driven risk-averse control strategy for stochastic nonlinear systems. The presented controller consists of two components: 1) A gain-scheduling control to ensure the invariance of the safe set by leveraging the concept of λ-contractivity, and 2) a nonlinear control to mitigate or eliminate the impact of the system’s nonlinearities on set invariance. To formalize a data-driven approach, closed-loop data-driven representations of both the gain scheduling and nonlinear dynamics are provided. These representations are then leveraged to impose λ-contractivity using a semidefinite programming (SDP) problem. The decision variables are optimized to minimize the variance of the system’s state distribution, ensuring that the system remains within the safe set with a specified probability. Additional constraints are incorporated to ensure robustness against unmeasured noise and residual nonlinearities. Simulation result validates the proposed approach, demonstrating its potential to enhance both safety and performance in stochastic nonlinear systems by managing gain scheduling, nonlinear dynamics, and uncertainty in a risk-aware manner.

Keywords: Robotics, Estimation, Biologically-inspired methods
Abstract: In a multi-modal system which combines thruster and legged locomotion such our state-of-the-art Harpy platform to perform dynamic locomotion. Therefore, it is very important to have a proper estimate of Thruster force. Harpy is a bipedal robot capable of legged-aerial locomotion using its legs and thrusters attached to its main frame. we can characterize thruster force using a thrust stand but it generally does not account for working conditions such as battery voltage. In this study, we present a momentum-based thruster force estimator. One of the key information required to estimate is terrain information. we show estimation results with and without terrain knowledge. In this work, we derive a conjugate momentum thruster force estimator and implement it on a numerical simulator that uses thruster force to perform thruster-assisted walking.

Keywords: Autonomous robots, Automotive control, Predictive control for nonlinear systems
Abstract: This paper presents a self-driving controller for electric wheelchairs designed to navigate narrow, winding paths while safely accepting potential soft collisions with pedestrians. Electric wheelchairs in urban areas often encounter challenges, known as the freezing robot problem, when maneuvering through tight spaces with pedestrians due to uncertain traffic participants' behavior. This study aims to develop an autonomous electric wheelchair capable of operating in dynamic and constrained environments by leveraging model predictive control specifically tailored for congested environments. To ensure real-time capability, the nonlinear dynamics are transformed into equivalent two linear dynamics of steering and velocity through a time-state control form that utilizes the travel distance along the curved path. A linear model predictive steering control is implemented to prevent collision with traffic participants, including pedestrians, while a stochastic model predictive velocity control suppresses the expected value of relative velocity to reduce the potential impact of collisions. Testing in narrow and curved paths with traffic participants using a LiDAR-equipped electric wheelchair has demonstrated the safety and practicality in low-speed locomotion.

Keywords: Robotics, Lyapunov methods, Constrained control
Abstract: This letter deals with enforcing safety in position-controlled robotic systems. We propose a safety-critical position control framework that enables a robot to safely avoid unsafe reference input positions without compromising the overall tracking performance of the existing controller. Simulation results, as well as experimental results on a quadrupedal robot, validate the practicality of our proposed controller for position-controlled robots. Our proposed approach, which uses the well-known Control Barrier Functions (CBF) framework in a model-free fashion, can serve as a safety filter alongside any model-based or model-free controller.

Keywords: Vision-based control, Adaptive systems, Visual servo control
Abstract: This paper presents an adaptive kinematic controller for quadrotors within the image-based visual servoing (IBVS) framework to track a planar, non-holonomic target with constant linear body fixed and arbitrary heading velocity. State-of-the-art image moment-based features in a virtual image plane are utilized to decouple translational and rotational kinematics. We formulate a novel adaptive control problem while designing the translational kinematic controller based on target heading reconstruction, ensuring exponential convergence of feature tracking and parameter estimation error to zero under the less restrictive initial excitation (IE) condition. Further, we design a heading controller augmented with a target heading velocity estimator. High-fidelity simulations demonstrate the controller's superior feature tracking performance compared to existing methods.

Keywords: Robotics, Human-in-the-loop control, Delay systems
Abstract: In this paper, it will be demonstrated that the classical scattering transformation is in fact a neural network. This insight is helpful in the sense that it connects the classical scattering transformation theory with the neural network field. By using this principle, we can deductively employ the knowledge suggested by the neural network theory to design new scattering transformation-based communication laws that can resolve several technical problems prevailing in the telerobotics field. As an example, the total number of neurons of the classical scattering transformation is increased to enable independent contact force feedback to enhance teleoperation transparency. A simulation result is provided.

Keywords: Human-in-the-loop control, Robotics
Abstract: The study addresses the challenge of enhancing human comfort during human-robot collaboration (HRC) tasks, where misalignment between robot behaviors and human comfort preferences is common. Current approaches often fail to adapt robot behavior effectively according to human preferences, leading to discomfort and less satisfying collaboration experience. To address this limitation, a multi-linear regression-based comfort prediction model is proposed to fine-tune robot behavioral factors and improve human comfort responses. In the experiment, three HRC object-delivery tasks were configured with high, medium, and low comfort levels based on different robot behaviors. Five behavioral factors—delivery distance, speed, height, trajectory, and gripper pose were varied, and participants rated their comfort on a five-point Likert scale. The results showed that the overall and factor-based comfort ratings largely aligned with the expected comfort levels of the tasks, indicating the model’s reliability in predicting and enhancing human comfort.

Keywords: Biological systems, Linear systems, Emerging control applications
Abstract: Animal sensorimotor behavior is frequently modeled using optimal controllers. However, it is unclear how the neural circuits within the animal's nervous system implement optimal controller-like behavior. In this work, we study the question of implementing a delayed linear quadratic regulator with linear dynamical "neurons" on a muscle model. We show that for any second-order controller, there are three minimal neural circuit configurations that implement the same controller. Furthermore, the firing rate characteristics of each circuit can vary drastically, even as the overall controller behavior is preserved. Along the way, we introduce concepts that bridge controller realizations to neural implementations that are compatible with known neuronal delay structures.

Keywords: Lyapunov methods, Reduced order modeling, Hybrid systems
Abstract: This paper presents a nonlinear feedback control design to stabilize the configuration of two swimmers moving inline in a uniform flow. We use a control-theoretic analysis of an experimentally validated model of inline swimming to characterize the behavior and stable points of the open-loop system. We find a linear control law that manipulates the flapping phase offset to stabilize the swimmers to a desired inline separation distance. We identify an idealized Hamiltonian version of the system and use it to find a nonlinear control law to improve the convergence rate of the linear control law. A combination of the two control laws stabilizes the system to a specified configuration faster than the natural dynamics.

Keywords: Control of networks, Biologically-inspired methods, Stability of nonlinear systems
Abstract: Multi-agent systems in biology, society, and engineering are capable of making decisions through the dynamic interaction of their elements. Nonlinearity of the interactions is key for the speed, robustness, and flexibility of multi-agent decision-making. In this work we introduce modulatory, that is, multiplicative, in contrast to additive, interactions in a nonlinear opinion dynamics model of fast-and-flexible decision-making. The original model is nonlinear because network interactions, although additive, are saturated. Modulatory interactions introduce an extra source of nonlinearity that greatly enriches the model decision-making behavior in a mathematically tractable way. Modulatory interactions are widespread in both biological and social decision-making networks; our model provides new tools to understand the role of these interactions in networked decision-making and to engineer them in artificial systems.

Keywords: Biologically-inspired methods, Robust control, Uncertain systems
Abstract: Nonlinearity and uncertainty are major features in control systems. In this context, the present work proposes to merge the brain emotional learning model with the benefits of robust event-driven control to handle uncertain nonlinear systems. The state-dependent unmodeled dynamics is estimated via the limbic system-inspired learning algorithm and added to the nominal control signal for compensation purposes. Furthermore, aiming at reducing data processing, and inherently, computational cost, the controller is triggered asynchronously driven by events function. Moreover, the closed-loop stability of the proposed control scheme is verified through the Lyapunov formalism, as well as the sampling admissibility to prevent the Zeno phenomena. The performance observed in the numerical results witnesses the effectiveness of the proposed control scheme.

Keywords: Biomedical, Robust control, Control applications
Abstract: This paper presents a just-in-time adaptive message intervention framework to promote long-term physical activity in young adults with insufficient activity levels. The framework personalizes interventions by approximating individual activity patterns using real-time smartwatch data. Then, a Risk-Sensitive Shrinking Horizon Model Predictive Control (MPC) is employed and reformulated as a Mixed-Integer Linear Programming (MILP) problem to optimize intervention design. To enable real-time adaptive decision-making on smartphones with limited computational resources, a neural network is trained offline using MILP-generated data, providing an efficient online control policy. Results from TryAIM clinical trial indicate that individuals respond differently to interventions, and for many, the models suggest that selecting the right time and message can effectively enhance physical activity levels.

Keywords: Autonomous robots, Biologically-inspired methods, Optimal control
Abstract: Bimanual tasks performed by human agents present unique optimal control considerations compared to cyberphysical agents. These considerations include minimizing attention, distributing that attention across two isolated hands, and coordinating the two hands to reach a broader goal. In this work, we propose a two-layer optimization problem that explicitly quantifies these considerations. The upper layer solves an attention distribution problem, while the two lower layer controllers (one per hand) tracks a trajectory using the upper layer solution. We introduce a formulation for attention control, where attention is a vector that is bound within a hyperbolic feasible region, determined by specifications of the lower layer controllers. We use this two-layer controller to optimize a single-player game of pong, rallying the ball between two paddles for as long as possible. We find several emergent behaviors from this optimization in simulations; stronger coordination leads to lower long-term attention, while high asymmetry and aggressive centering for stability increase overall attention.

Keywords: Stochastic optimal control, Constrained control, Robotics
Abstract: This paper addresses the control problem of mobile robots with random noises under timed reach-avoid (TRA) tasks. TRA tasks are expressed as signal temporal logic (STL) formulas, and an optimization problem (OP) is formulated such that the chance constraint (CC) is embedded. To deal with the OP in the continuous-time setting, a local-to-global control strategy is proposed. We first decompose the STL formula into a finite number of local ones, and then decompose and convert the CC into deterministic constraints such that a finite number of local OPs are established and solved efficiently. The feasibility of all the local OPs implies the feasibility of the original OP, which results in a control strategy for the task accomplishment. The proposed strategy is further extended to the multi-robot case. Finally, numerical examples and comparisons are presented to illustrate the efficacy of the proposed control strategy.

Keywords: Stochastic optimal control, Switched systems, Uncertain systems
Abstract: The contribution of this paper is the stabilization in the mean-square sense of discrete-time Markov jump linear systems with mixed known, unknown, and time-varying transition probabilities. To handle uncertainties in the transition probabilities, we develop a control strategy utilizing mode-dependent stationary state feedback controllers and introduce data-based ambiguity sets that, extending existing literature, account for known, unknown and time-varying probabilities. These ambiguity sets are constructed using estimated transition matrices and probabilistic bounds derived from the Dvoretzky-Kiefer-Wolfowitz inequality. We validate the effectiveness of our method with numerical simulations on a control system subject to deadline overruns, demonstrating the improvements of incorporating partial knowledge of the transition probabilities.

Keywords: Stochastic systems, Agents-based systems, Modeling
Abstract: Recommendation systems underlie a variety of online platforms. These recommendation systems and their users form a feedback loop, wherein the former aims to maximize user engagement through personalization and the promotion of popular content, while the recommendations shape users’ opinions or behaviors, potentially influencing future recommendations. These dynamics have been shown to lead to shifts in users’ opinions. In this paper, we ask whether reactive users, who are cognizant of the influence of the content they consume, can prevent such changes by actively choosing whether to engage with recommended content. We first model the feedback loop between reactive users’ opinion dynamics and a recommendation system. We study these dynamics under three different policies - fixed content consumption (a passive policy), and decreasing or adaptive decreasing content consumption (reactive policies). We analytically show how reactive policies can help users effectively prevent or restrict undesirable opinion shifts, while still deriving utility from consuming content on the platform. We validate and illustrate our theoretical findings through numerical experiments.

Keywords: Stochastic optimal control, Hierarchical control, Autonomous systems
Abstract: This paper addresses the problem of hierarchical task control, where a robotic system must perform multiple subtasks with varying levels of priority. A commonly used approach for hierarchical control is the null-space projection technique, which ensures that higher-priority tasks are executed without interference from lower-priority ones. While effective, the state-of-the-art implementations of this method rely on low-level controllers, such as PID controllers, which can be prone to suboptimal solutions in complex tasks. This paper presents a novel framework for hierarchical task control, integrating the null-space projection technique with the path integral control method. Our approach leverages Monte Carlo simulations for real-time computation of optimal control inputs, allowing for the seamless integration of simpler PID-like controllers with a more sophisticated optimal control technique. Through simulation studies, we demonstrate the effectiveness of this combined approach, showing how it overcomes the limitations of traditional methods by optimizing the task performance.

Keywords: Stochastic optimal control, Constrained control, Randomized algorithms
Abstract: Dynamic Programming suffers from the well-known ``curse of dimensionality'', further exacerbated by expectations in stochastic systems. This paper presents a Monte Carlo-based sampling approach of the state and input spaces and an interpolation procedure for the resulting value function in a "self-approximating" fashion, eliminating the need for ordering or set-membership tests. We provide a proof of almost sure convergence for the value iteration (and consequently, policy iteration) procedure. The proposed sampling and self-approximating algorithm alleviates the burden of gridding and interpolation traditionally required in DP. Moreover, we demonstrate that the proposed interpolation procedure is well-suited for handling probabilistic constraints by sampling both infeasible and feasible regions. The curse of dimensionality cannot be avoided, however, this approach offers a convenient framework for addressing lower-order nonlinear stochastic systems with probabilistic constraints.

Keywords: Control over communications, Markov processes, Robust control
Abstract: This letter addresses a fundamental issue of time-varying parametric uncertainty affecting unreliable communication links that convey the control commands to actuators in wireless networked control systems. It introduces the polytopic time-inhomogeneous finite-state Markov channel model to account for significant changes in possible propagation channel characteristics and analytically solves the linear quadratic regulation problem under the generalized packet dropout compensation. An example validating the results is presented.

Keywords: Time-varying systems, Reduced order modeling, Simulation
Abstract: Physics-based electrochemical models offer great potential for battery management systems by accurately capturing internal states of the battery. However, their strong nonlinearity and coupled partial differential equations result in high computational complexity, limiting real-time applications. Model order reduction techniques address this challenge, yet traditional methods, such as Padé approximation ´ and polynomial-based approaches, assume time-invariant parameters, leading to errors under dynamic conditions. This study systematically evaluates the representatively model order reduction techniques, highlighting their limitations in handling time-varying battery dynamics. To overcome these challenges, a novel order-reduction approach is proposed, integrating proper orthogonal decomposition with the finite difference method to improve computational efficiency and accuracy. Simulation results demonstrate that the proposed method significantly reduces computational cost while maintaining high accuracy. Experimental validation further confirms its robustness, establishing it as a promising candidate for real-time battery management system applications.

Keywords: Constrained control, Control applications, Predictive control for nonlinear systems
Abstract: Control barrier functions (CBFs) have gained attention as a method for ensuring the safety of dynamic systems by enforcing constraints on the time derivative of a value function. Ordinarily used for trajectory control of mechanical systems, this paper proposes a CBF-based approach for the fast and safe charging of a lithium-ion battery cell under temperature, charge current, and terminal voltage constraints. The CBF approach is compared to model predictive control (MPC), which has recently emerged as a promising control strategy for battery management due mainly to its capability for real-time constraint handling. Comparative simulation results illustrate key advantages and trade-offs between the two approaches and show that CBFs can serve as a viable alternative to MPC for safe fast-charge applications.

Keywords: Observers for nonlinear systems, Estimation
Abstract: Lithium-ion batteries are critical for modern energy storage systems, with accurate modeling essential for optimized performance, safety, and operational life. This paper investigates how the choice of spatial discretization method for the widely-used, partial differential equation-based, single-particle model affects the nonlinear observability properties of electrochemical model-based observers. Despite different spatial discretization schemes achieving similar voltage prediction accuracy with respect to experimental data, the choice of scheme leads to state-space representations that have different observability properties. We also find that increasing the number of states of the internal model decreases the expected quality of the state estimates, as the newly introduced states tend to be nearly redundant copies of existing states. These findings emphasize the importance of balancing model complexity, experimental accuracy, and nonlinear observability—an often-overlooked aspect—when selecting electrochemical models for observer design. Greater attention to nonlinear observability could lead to more effective and reliable battery management systems.

Keywords: Modeling, Model Validation, Identification
Abstract: Electrode open-circuit potential (OCP) is an intrinsic electrochemical property unique to the electrode material within a lithium-ion battery. It is critical to develop models of electrode OCP for accurate physics-based cell models used in modeling, estimation, and control applications. Measuring this relationship directly requires cell teardown, a costly procedure that destroys the cell. It is beneficial to develop alternative methods to estimate OCP avoiding teardown. This paper develops a series of methods to estimate electrode OCP using only cell open-circuit voltage (OCV) tests and a literature estimate of a known electrode OCP. We develop two main methods which were implemented in MATLAB. The first estimates the positive-electrode OCP using a given multi-species multi-reaction (MSMR) model of the negative electrode's OCP. The second estimates the negative-electrode OCP assuming an MSMR model of the positive-electrode's OCP with a separate procedure developed for LFP//graphite cells. Results were compared to measured cell OCV and electrode OCP. While these methods are heuristic, attempts to development mathematical algorithms to solve this problem have fallen short due to fundamental unobservability. We conclude that these methods might suit an AI application. These methods demonstrate that it is possible to find an estimate of electrode OCP without cell teardown as long as at least one of the electrode materials is known.

Keywords: Process Control, Predictive control for nonlinear systems, Learning
Abstract: Model predictive control (MPC) is a powerful tool for controlling complex nonlinear systems under constraints, but often struggles with model uncertainties and the design of suitable cost functions. To address these challenges, we propose an approach integrating MPC with safe Bayesian optimization to improve long-term closed-loop performance despite significant model-plant mismatches. By parameterizing the MPC stage cost function using a radial basis function network, we employ Bayesian optimization as a multi-episode learning strategy to tune the controller without relying on precise system models. This method mitigates conservativeness introduced by overly cautious soft constraints in the MPC cost function and provides probabilistic safety guarantees during learning, ensuring that safety-critical constraints are met with high probability. As a practical application, we apply our approach to fast charging of lithium-ion batteries, a challenging task due to the complicated battery dynamics and strict safety requirements, subject to the requirement to be implementable in real time. Simulation results show that our method reduces charging times compared to traditional MPC approaches while maintaining safety, even in the presence of model-plant mismatch.

Keywords: Power systems, Optimization, Machine learning
Abstract: Battery energy storage systems (BESS) play an increasingly vital role in integrating renewable generation into power grids due to their ability to dynamically balance supply. Grid-tied batteries typically employ power converters, where part-load efficiencies vary non-linearly. While this non-linearity can be modeled with high accuracy, it poses challenges for optimization, particularly in ensuring computational tractability. In this paper, we consider a non-linear BESS formulation based on the Energy Reservoir Model (ERM). A data-driven approach is introduced with the input-convex neural network (ICNN) to approximate the nonlinear efficiency with a convex function. The epigraph of the convex function is used to engender a convex program for battery ERM optimization. This relaxed ICNN method is applied to a battery revenue maximization problem and is compared with three other ERM formulations (nonlinear, linear and mixed-integer). Specifically, ICNN-based method appear to be promising for future battery optimization with desirable feasibility and optimality outcomes across revenue maximization use-case.

Keywords: Predictive control for nonlinear systems, Linear parameter-varying systems, Energy systems
Abstract: Advancements in wind turbine technology have made wind energy more cost-competitive. While taller towers use less material, they are more susceptible to fatigue. This letter introduces a convex model predictive control scheme to actively counteract side-side periodic loads using a velocity-based approach, which captures the system’s nonlinear behavior without requiring extensive prior operating points. A quasi-linear parameter-varying dynamic model for wind turbine towers is established through model demodulation transformation. Simulation results show a 96% reduction in net force in the side-side direction at the tower top under turbulent wind conditions.

Keywords: Adaptive control, Energy systems, Flexible structures
Abstract: Wind turbines are getting larger to increase power capacity. Their longer blades sample a larger area of the spatially and temporally varying turbulent wind field, leading to increased periodic blade load and fatigue damage over time. Individual pitch control (IPC) has proven effective in alleviating these loads by pitching the blades. Conventional IPC fully attenuates the periodic blade loads, which requires excessive pitching, leading to additional stresses on the pitch system. To balance pitch actuation and load alleviation, bounds can be set on the pitch signal (input-constrained IPC), or on the load (output-constrained IPC). While input-constrained IPC has been abundantly researched, little research has focused on output-constrained IPC and on the trade-off when operating between full IPC and no IPC. Therefore, we propose an output-constrained IPC method using an adaptive leaky integrator. The natural frequency of the leaky integrator is adapted on the error between the reference and resultant blade moment. This allows the control scheme to attain every load alleviation level between full and no IPC. Furthermore, in realistic turbulent wind conditions, operating close to full IPC leads to diminishing returns, showing that the proposed controller achieves a superior trade-off between load reduction and actuator effort.

Keywords: Predictive control for linear systems, Mechanical systems/robotics, Stochastic optimal control
Abstract: The literature on stochastic MPC (SMPC) claims probabilistic constraints give a major feasibility problem and suggest solutions based on a combination of feedback and MPC. This paper explains it makes sense to relax the probabilistic constraints after the first part of the prediction horizon which solves the feasibility problem. Further, it is demonstrated that the effect that is special to probabilistic constraints is that they change the distribution of inputs and outputs. Current SMPC uses the conditional constraint probability given current measurements as the design parameter. Here a method to relate to the more application relevant unconditional constraint probability is developed. Finally, all the points are demonstrated with a simple wind turbine control problem which served as the motivation for this research in the first place.

Keywords: Energy systems, Power systems, Decentralized control
Abstract: As wind turbine sizes and their rated power capacities increase, the spatial and temporal load imbalances over the rotor surface increase due to larger wind asymmetries, aerodynamic imbalances, and calibration offsets. The multi-blade coordinate (MBC) transform-based individual pitch control (IPC) has garnered significant attention in the literature, and it considers loads in a non-rotating reference frame. This leads to increased pitch actuation for all wind turbine blades when subject to large load imbalances. On the contrary, the single-blade control (SBC) IPC strategy is well suited for handling such load imbalances as it involves equipping each blade with a localized control system, thereby ensuring an independent operation in a rotating reference frame. However, unlike MBC transform-based IPC, the effects of system phase lag on SBC performance have not been investigated. This article investigates the effects of such phase lags and multivariable coupling on SBC performance, and proposes a novel framework for phase compensation in SBC as a convenient method for constructing and calibrating a lead compensator. Using mid-fidelity OpenFAST simulations, it is demonstrated that phase compensation in SBC improves load mitigation at the targeted frequencies and reduces actuation effort. In contrast, the absence of phase compensation can lead to load amplifications, especially for larger wind turbines.

Keywords: Energy systems, Control applications, Emerging control applications
Abstract: Denser turbine spacing in wind farms leads to increased wake interactions, causing power losses when each turbine operates under its own greedy control scheme. To mitigate these effects, research is exploring strategies that consider the entire wind farm rather than singular turbines. The so-called helix approach has recently gotten significant attention from the research community. It aims to reduce wake losses through periodic individual pitch control. Wake steering on the other hand uses yaw actuation to laterally deflect the wake away from downstream turbines. In this paper, we adapt and validate a steady-state surrogate model to compute the time-averaged velocity field behind a wind turbine operating with the helix approach. The model is tuned using data from Large Eddy Simulations. We compare the helix model to wake steering and baseline operation in a wind farm case study, demonstrating that the helix approach offers promising benefits under specific wind conditions.

Keywords: Observers for nonlinear systems, Kalman filtering, Estimation
Abstract: In the field of wind farm flow control and monitoring, most studies rely on steady-state engineering models. Most of these models are based on a statistical description of wakes and assume uniform wind conditions across the farm. They take as inputs at least the free-flow wind conditions, possibly the yaw misalignment angles between the rotor axis, wind direction and some parameters of the wake models, which can also vary. This work introduces an application of Ensemble Kalman Filters, where they are used to estimate the inputs required by a steady-state wind farm solver to align with a set of uncertain or biased measurements from a wind farm SCADA system. While the primary objective is estimating the free-flow wind conditions, strategies are proposed to also estimate biases in the nacelle and wind vane position measurements, potentially enabling their detection and compensation without the need for additional sensors. The performance of the EnKFs is evaluated using both synthetic and real SCADA data, with heuristic free-flow wind estimators serving as benchmarks.

Keywords: Game theory, Randomized algorithms, Robust control
Abstract: A scenario-based risk-sensitive optimization framework is presented to approximate minimax solutions with high confidence. The approach involves first drawing several random samples from the maximizing variable, then solving a sample-based risk-sensitive optimization problem. This work derives the sample complexity and the required risk-sensitivity level to ensure a specified tolerance and confidence in approximating the minimax solution. The derived sample complexity highlights the impact of the underlying probability distribution of the random samples. The framework is demonstrated through applications to zero-sum games and model predictive control for linear dynamical systems with bounded disturbances.

Keywords: Game theory, Stability of nonlinear systems, Stability of linear systems
Abstract: Definition of the (weak) Pareto improvement and the trap of weak Pareto improvement are given for noncooperative dynamical systems with vector-valued payoff functions. Specifically, we develop an incentive mechanism that satisfies sustainable budget constraint. We assume that there is a system manager who is authorized to design the incentive functions in order to maximize the social welfare of her choice. Specifically, the socially maximum state is predicated on the weighted sum of all the payoff functions. It turns out that depending on the choice of the design parameters, we may observe that some of the agents can be trapped in a weak Pareto improvement even though the system trajectory is weakly Pareto improving. Our results is a generalized version of the results for scalar-valued payoff functions.

Keywords: Game theory, Stochastic systems, Stability of nonlinear systems
Abstract: Markov games provide a powerful framework for modeling strategic multi-agent interactions in dynamic environments. Traditionally, convergence properties of decentralized learning algorithms in these settings have been established only for special cases, such as Markov zero-sum and potential games, which do not fully capture real-world interactions. In this paper, we address this gap by studying the asymptotic properties of learning algorithms in general-sum Markov games. In particular, we focus on a decentralized algorithm where each agent adopts an actor-critic learning dynamic with asynchronous step sizes. This decentralized approach enables agents to operate independently, without requiring knowledge of others' strategies or payoffs. We introduce the concept of a Markov Near-Potential Function (MNPF) and demonstrate that it serves as an approximate Lyapunov function for the policy updates in the decentralized learning dynamics, which allows us to {characterize the convergent set of strategies}. We further strengthen our result under specific regularity conditions and with finite Nash equilibria.

Keywords: Game theory, Uncertain systems, Optimization
Abstract: In this work, we present linearized probabilistic weapon engagement zones (linearized PEZ), which provide a method to prevent agents from engaging in a pursuer-evasion differential game while accounting for uncertainty in the parameters of the game. This type of differential game is commonly used to model adversarial environments, where the parameters of the adversary (pursuer), such as initial positions and range, are often unknown or uncertain. Additionally, with the increasing potential of GPS jamming in modern warfare, even friendly (evader) parameters, such as position and orientation, can be uncertain. We demonstrate that linearized PEZ effectively approximates the true engagement zone distribution found using a Monte Carlo method. Using linearized PEZ, we develop a path optimization algorithm that plans safe trajectories while addressing this uncertainty. Numerical simulations are presented to demonstrate the effectiveness of our approach.

Keywords: Game theory
Abstract: We consider multi-agent systems with general information networks where an agent may only observe a subset of other agents. A system designer assigns local utility functions to the agents guiding their actions towards an outcome which determines the value of a given system objective. The aim is to design these local utility functions such that the Price of Anarchy (PoA), which equals the ratio of system objective at worst possible outcome to that at the optimal, is maximized. Towards this, we first develop a linear program (LP) that characterizes the PoA for any utility design and any information network. This leads to another LP that optimizes the PoA and derives the optimal utility design. Our work substantially generalizes existing approaches to the utility design problem.

Keywords: Optimization, Game theory, Networked control systems
Abstract: In this paper, we study the existence of a Quasi Nash Equilibrium (QNE) for network games with low-level of interaction among the players. Using our previous results on the existence of solutions for non-monotone variational inequalities (VIs), we obtain new sufficient conditions for the existence of a QNE. The conditions are captured by lower bounds on the interaction graph and influence matrix capturing the effects of players decisions and their interactions. We also consider a general game under a P-property of the game Jacobian, and provide a new sufficient condition for the existence of a QNE.

Keywords: Numerical algorithms, Optimization algorithms, Aerospace
Abstract: The turbine blade is one of the most critical components of a spacecraft. To enhance the efficiency of batch inspection of turbine blades, this paper proposes a fully automatic slice feature extraction method for 3D measurement of turbine blades based on point cloud. The algorithm uses a non-iterative hierarchical clustering method to quickly identify the blade body, followed by a convex hull-based search for the main feature lines and automatic cross-section parameter establishment. A curve trajectory projection method is introduced to improve accuracy by fitting a local quadratic surface and curving the projection, avoiding the extrusion effect of vertical projection. Experimental results confirm that the method enhances the efficiency of turbine blade inspection.

Keywords: Control applications, Mechatronics, Control laboratories
Abstract: Feedback control has benefited society in many ways, but in order to fully realize the potential benefits, feedback control needs to be implemented in a wider range of physical systems and products. Conducting real-time feedback control experiments can be difficult and expensive. This work seeks to make it easier for controls engineers to get started with real-time experiments using low-cost hardware and open-source software. A Raspberry Pi mini-computer can be a cost-effective option for performing feedback control experiments, but programming it for real-time execution of a control law can be complicated. This paper makes two contributions to advancing broader use of Raspberry Pi's in feedback control implementation: 1) verification of the ability of the Raspberry Pi to consistently execute the control law at hard real-time intervals on the order of one to two milliseconds and 2) an object-oriented C++ library to make it easier to write real-time code that mimics a classical controls block diagram

Keywords: Adaptive control, Manufacturing systems and automation, Control applications
Abstract: This paper provides a novel approach to controlling the melt pool (MP) height and temperature in directed energy deposition (DED) additive manufacturing, addressing the challenge of laser efficiency uncertainty caused by variations in laser power and material absorption characteristics, which affect the quality and consistency of produced parts. Nonlinear governing equations for MP height and temperature are derived and a state space formulation is provided that includes uncertainty in laser efficiency and other process parameters. A stable indirect adaptive controller is designed by employing a nondimensional, scaled form of the governing equations for process operation around an operating point and an adaptive estimation algorithm based on the full nonlinear dynamics. Since the uncertainty in the laser efficiency is typically bounded in a range around a nominal value, a smooth-parameter projection adaptation algorithm is employed. The approach also allows us to estimate other coupled uncertain material and system parameters. Extensive numerical simulations were conducted with conditions relevant to the practical application to evaluate the proposed adaptive controller, and a representative sample of the results is presented and discussed.

Keywords: Iterative learning control, Manufacturing systems
Abstract: This paper proposes an Iterative Learning Control (ILC) algorithm to determine the laser power input during the deposition of each layer, in order to regulate the melt pool area (MPA) in the Directed Energy Deposition (DED) metal additive manufacturing process. Based on the laser power applied and the corresponding MPA error obtained experimentally, the ILC algorithm updates the laser power so that the MPA error is reduced in the next iteration. It turns out that, even with the well-tuned ILC algorithm, the MPA error persists in lower layers due to the heat dissipation properties of the cold substrate. As a remedy of this issue, the substrate pre-heating strategy is introduced. The combination of the ILC algorithm with the substrate pre-heating shows the potential in minimizing the MPA error across all the layers of the deposited part, thereby enhancing the quality in DED process.

Keywords: Iterative learning control, Mechatronics, Optimization
Abstract: Iterative learning control (ILC) techniques are capable of improving the tracking performance of control systems that repeatedly perform similar tasks by utilizing data from past iterations. The aim of this paper is to design a systematic approach for learning parameterized feedforward signals with limited complexity. The developed method involves an iterative learning control in conjunction with a data-driven sparse subset selection procedure for basis function selection. The ILC algorithm that employs sparse optimization is able to automatically select relevant basis functions and is validated on an industrial flatbed printer.

Keywords: Networked control systems, Cooperative control, Game theory
Abstract: We propose a protocol based on mechanism design theory and encrypted control to solve average consensus problems among rational and strategic agents while preserving their privacy. The proposed protocol provides a mechanism that incentivizes the agents to faithfully implement the intended behavior specified in the protocol. Furthermore, the protocol runs over encrypted data using homomorphic encryption and secret sharing to protect the privacy of agents. We also analyze the security of the proposed protocol using a simulation paradigm in secure multi-party computation. The proposed protocol demonstrates that mechanism design and encrypted control can complement each other to achieve security under rational adversaries.

Keywords: Agents-based systems, Network analysis and control, Control of networks
Abstract: This work extends the recent opinion dynamics model from (Cheng et al., 2019), emphasizing the role of group pressure in consensus formation. We generalize the findings to incorporate social influence algorithms with general time-varying, opinion- dependent weights and multidimensional opinions, beyond bounded confidence dynamics. We demonstrate that, with uni- formly positive conformity levels, group pressure consistently drives consensus and provide a tighter estimate for the conver- gence rate. Unlike previous models, the common public opinion in our framework can assume arbitrary forms within the convex hull of current opinions, offering flexibility applicable to real- world scenarios such as opinion polls with random participant selection. This analysis provides deeper insights into how group pressure mechanisms foster consensus under diverse conditions.

Keywords: Optimization algorithms, Optimization
Abstract: The paper proposes the Consensus Augmented Lagrange Alternating Direction Inexact Newton (Consensus ALADIN) algorithm, a novel approach for solving distributed consensus optimization problems (DC). Consensus ALADIN allows each agent to independently solve its own nonlinear programming problem while coordinating with other agents by solving a consensus quadratic programming (QP) problem. Building on this, we propose Broyden–Fletcher–Goldfarb–Shanno (BFGS) Consensus ALADIN, a communication-and-computation-efficient Consensus ALADIN. BFGS Consensus ALADIN improves communication efficienc through BFGS approximation techniques and enhances computational efficiency by deriving a closed form for the consensus QP problem. Additionally, by replacing the BFGS approximation with a scaled identity matrix, we develop Reduced Consensus ALADIN, a more computationally efficient variant. We establish the convergence theory for Consensus ALADIN and demonstrate its effectiveness through application to a nonconvex sensor allocation problem.

Keywords: Power systems, Distributed control, Networked control systems
Abstract: The frequency synchronization problem of wide-area control from the higher level of networked power systems is considered. We investigate the exponential convergence and robustness of the passivity-based distributed averaging controller used in damping the inter-area oscillations of power networks and coping with the time-varying power demand. A strict composite storage function is constructed that allows us to quantify the exponential convergence rate of the closed-loop system. Furthermore, L2 gain performance is utilized as a tool to quantify the impact of time-varying power demand on control areas. We simulate the controller performance on a four-area network which is equivalent to the IEEE New England 39-Bus system.

Keywords: Modeling, Networked control systems, Network analysis and control
Abstract: We examine an effective dynamic reconfiguration model for triangulated origami structures using planar straight-line graphs. In this setup, the origami panels are first represented as edges and vertices in an undirected triangular graph. A triangulated consensus protocol for the corresponding origami formation control problem is then developed, where state of the nodes reach agreement on target configuration while ensuring that the reconfiguration process is realizable by each panel--the setup is then extended to the entire triangulated origami structure. The proposed approach provides a general graph-theoretic framework for expressing the geometric evolution of diverse origami patterns during the folding/unfolding process.

Keywords: Agents-based systems, Cooperative control, Autonomous systems
Abstract: In this work a novel approach to containment control for multi-agent systems is presented, leveraging techniques from opinion dynamics and considering a scenario where agents modeled as double integrators are taken into account. The defined control law is specifically capable of guiding follower agents within the convex hull of leader agents, taking into account and managing the presence of uncertainties in the available information. In particular, the technique is based on the design of virtual leaders that can safely guide the follower agents into the convex hull of the real leaders. Various design strategies for these virtual leaders are presented, using the results obtained from opinion dynamics. The paper details the problem statement, assumptions, and the proposed control strategy, followed by numerical results validating the effectiveness of the approach.

Keywords: Nonlinear systems identification, Optimization, Simulation
Abstract: Driver heterogeneity is key to understanding traffic patterns and stochastic microsimulation. Previous studies have differentiated between inter-driver and intra-driver heterogeneity; this study focuses on inter-driver heterogeneity to represent the collective driver behavior within a population. Using the Intelligent Driver Model (IDM) and a standard nonlinear calibration technique, we found that the resulting parameter distributions often poorly reflect the actual inter-driver variability due to practical identifiability issues—multiple parameter sets fit the data equally well during calibration. We propose a centroid-guided calibration approach that reduces the spread of parameter estimates by eliminating the impact of unidentifiable parameters. Numerical experiments with synthetic and real-world data demonstrate that the proposed method more accurately captures inter-driver heterogeneity.

Keywords: Nonlinear systems identification, Automotive systems, Machine learning
Abstract: Accurate driver behavior modeling is essential for improving the interaction and cooperation of the human driver with the driver assistance system. This paper presents a novel approach for modeling the response of human drivers to visual cues provided by a speed advisory system using a Koopman-based method with online updates. The proposed method utilizes the Koopman operator to transform the nonlinear dynamics of driver-speed advisory system interactions into a linear framework, allowing for efficient real-time prediction. An online update mechanism based on Recursive Least Squares (RLS) is integrated into the Koopman-based model to ensure continuous adaptation to changes in driver behavior over time. The model is validated using data collected from a human-in-the-loop driving simulator, capturing diverse driver-specific trajectories. The results demonstrate that the offline learned Koopman-based model can closely predict driver behavior and its accuracy is further enhanced through an online update mechanism with the RLS method.

Keywords: Automotive systems, Sensor fusion, Control applications
Abstract: Accurate vehicle tracking is an important requirement on autonomous and semi-autonomous vehicles for safe operation. This paper presents a multi-stage high-gain observer formulation for vehicle tracking using various motion models. Vehicle tracking has previously relied on inaccurate relative motion models due to which tracking errors are obtained, especially during turning maneuvers of the ego vehicle. In this paper, the limitations of the previous motion models are delineated along with the development of new motion models. An important aspect of the new models is that they inherently need ego state estimates to track the other vehicles on the road. The previously developed relative models neglect the motion of the ego vehicle. In the first newly developed model the ego state estimates are accounted for in the measurement equation while in a second model, they are accounted for in the process dynamics. Furthermore, in this paper the new motion models are transformed into companion form, and a high-gain observer formulation is then provided for each model along with ego state estimation. Such a multi-stage high gain observer performs simultaneous ego state estimation and vehicle tracking for both inertial and relative motion models. The performance of this multi-stage high-gain observer is validated using experimental data from an autonomous vehicle at the University of Minnesota.

Keywords: Modeling, Computational methods, Automotive systems
Abstract: Aircraft tire-runway interactions is critical for safe operation of aircraft at landing and take-off phases. Transverse grooves are built on airport pavement surface to help drain any possible accumulated water and mitigate aircraft overrun due to reduced friction coefficient on raining conditions. It is unclear how the grooving runway configurations can maintain and even increase the skid resistance. We present a modeling and experimental study to evaluate tire-runway friction on wet, grooved pavement surfaces. The model captures the tire-runway interlock force that contributes to the total tire-runway friction. The model also incorporates the effect of the drainage capacity of the grooves on the tire-pavement contact patch size. The total friction force is obtained with the integrated interlock effect and the LuGre friction model. Comparisons between the model prediction and the experimental results from an indoor tire-runway testbed are presented to validate and demonstrate the proposed friction force model.

Keywords: Reinforcement learning, Autonomous systems, Neural networks
Abstract: Reinforcement learning from human feedback (RLHF) has gained increasing attention in automated vehicle planning and control due to its potential to enhance decision-making processes and accelerate policy optimization. By incorporating human feedback into reinforcement learning models, RLHF enables agents to develop more reliable and context-aware behaviors, particularly in complex and dynamic traffic environments. This paper presents PVP with Auxiliary Network (aPVP), an RLHF-based framework designed to improve the stability of automated driving policies. Specifically, we extend the Proxy Value Propagation (PVP) framework by introducing an auxiliary neural network trained on human driving data. This auxiliary model serves as a virtual driver, providing a similarity-based loss function that guides the actor network to explore within a reasonable range, ensuring policy stability while preserving learning flexibility. To validate the effectiveness of aPVP, we design a comprehensive experimental setup. Empirical results demonstrate that the proposed approach significantly enhances policy stability compared to the original PVP framework. These findings highlight the potential of aPVP in improving RLHF-based decision-making systems, paving the way for future research on enhancing scalability and adaptability in real-world autonomous driving scenarios.

Keywords: Human-in-the-loop control, Automotive control, Identification
Abstract: This paper presents how game-theoretic level-k theory, a framework for modeling the cognitive decision-making level of agents with bounded rationality, can be used to describe the interactive decision-making policies within human-autonomy teams during collaborative tasks. The approach hypothesizes that the prediction of human behavior can be used to enable autonomous systems to make better decisions. A case study is used to create a preliminary data library mapping human behavior to specific level-k strategies. Specifically, a participant completed laps on a shared control driving simulator while performing a secondary task. Between laps, the participant is informed of changes in the vehicle controller that impact the influence of the human input on the vehicle motion, reflecting changes in the belief the human holds about the autonomy, or changing their level-k. Using the library, the mapping between human behavior and level-k theory is evaluated through the classification of interactions between human-autonomy teams under directed scenarios.

Keywords: Optimization, Optimization algorithms
Abstract: This letter details two novel algorithms for computing asset allocations given dynamic requests for support. Using spatial and temporal discretization, the first algorithm casts the allocation problem as a lexicographic integer-linear program (ILP) that is efficiently solved using an ILP solver. Improving computational efficiency, the second algorithm replaces the ILP solver with a novel dual-tactic linear program solver. Provided proofs, complexity analyses, and numerical analyses demonstrate the computational efficiency and convergence of both algorithms to performant solutions. Discussion of application to broad domains such as defense, conservation, and resource management for businesses is provided.

Keywords: Optimization, Randomized algorithms, Uncertain systems
Abstract: The scenario approach is widely used in robust control system design and chance-constrained optimization, maintaining convexity without requiring assumptions about the probability distribution of uncertain parameters. However, the approach can demand large sample sizes, making it intractable for safety-critical applications that require very low levels of constraint violation. To address this challenge, we propose a novel yet simple constraint scaling method, inspired by large deviations theory. Under mild nonparametric conditions on the underlying probability distribution, we show that our method yields an exponential reduction in sample size requirements for bilinear constraints with low violation levels compared to the classical approach, thereby significantly improving computational tractability. Numerical experiments on robust pole assignment problems support our theoretical findings.

Keywords: Optimization, Sensor networks, Variational methods
Abstract: Optimal sensing nodes selection (SNS) in dynamic systems is a combinatorial optimization problem that has been thoroughly studied in the recent literature. This problem can be formulated within the context of set optimization. For high-dimensional nonlinear systems, the problem is extremely difficult to solve. It scales poorly too. Current literature poses combinatorial submodular set optimization problems via maximizing observability performance metrics subject to matroid constraints. Such an approach is typically solved using greedy algorithms that require lower computational effort yet often yield sub-optimal solutions. In this paper, we address the SNS problem for nonlinear dynamical networks using a variational form of the system dynamics, that basically perturb the system physics. As a result, we show that the observability performance metrics under such system representation are indeed submodular. The optimal problem is then solved using the multilinear continuous extension. This extension offers a computationally scalable and approximate continuous relaxation with a performance guarantee. The effectiveness of the extended submodular program is studied and compared to greedy algorithms. We demonstrate the proposed set optimization formulation for SNS on nonlinear natural gas combustion networks.

Keywords: Optimization, Transportation networks
Abstract: This paper addresses the challenges of electric vehicles (EVs) long-distance mass evacuations, particularly those posed by the extended charging time. We focus on optimizing evacuation planning in high EV ownership areas by considering route selection, vehicle grouping, departure timing, and charging scheduling, while incorporating Mobile Charging Stations (MCS) to supplement the existing Fixed Charging Stations (FCS). A two-stage optimization approach is used, i.e. route optimization through a recursive Dijkstra algorithm, followed by vehicle scheduling and MCS deployment via Mixed Integer Linear Programming (MILP). Apart from demonstrating the effectiveness of MCS in reducing the evacuation time, the study reveals key insights on the optimal scheduling and MCS placement patterns. In addition, this paper also investigates the impact of nonuniform properties (such as battery sizes and initial energy level) among EVs on the evacuation time, relating to the more complicated real-world operating conditions. Furthermore, through quantifying the impact of infrastructure capacities on evacuation, specifically charging rate and traveling speed, insights on cost-effective resource allocation for infrastructure upgrade are generated. The outcome provides a valuable tool for local agencies to optimize and evaluate evacuation strategies and infrastructure, with results showcasing the significant potential of MCS in enhancing EV evacuations in high-risk regions.

Keywords: Optimization, Learning, Autonomous systems
Abstract: This paper introduces HyGIPO, a novel gradient-based iterative policy optimization technique designed for efficient policy learning in autonomous systems, especially in the presence of modeling errors. Performance of control algorithms for autonomous systems is often limited by mismatches between a simplified nominal model and a complex real system. To address this degradation, HyGIPO leverages a hybrid gradient optimization approach, combining gradients of dynamics from a nominal model with real-world data to optimize control policies. We apply this method to the quadcopter waypoint tracking problem, with the controller parameterized by a neural network, demonstrating its effectiveness in both simulation and hardware experiments. In simulation, HyGIPO rapidly learns the policy within a hundred samples, showing orders of magnitude higher sample efficiency compared to reinforcement learning methods. The hardware experiments further validate the method, achieving successful tracking results in just tens of samples.

Keywords: Optimization algorithms, Adaptive control, Uncertain systems
Abstract: Extremum seeking is an online, model-free optimization algorithm traditionally known for its practical stability. This paper introduces an extremum seeking algorithm designed for unbiased convergence to the extremum asymptotically, allowing users to define the convergence rate. Unlike conventional extremum seeking approaches utilizing constant gains, our algorithms employ time-varying parameters. These parameters reduce perturbation amplitudes towards zero in an asymptotic manner, while incorporating asymptotically growing controller gains. The stability analysis is based on state transformation, achieved through the multiplication of the input state by asymptotic growth function, and Lie bracket averaging applied to the transformed system. The averaging ensures the practical stability of the transformed system, which, in turn, leads to the asymptotic stability of the original system. Moreover, for strongly convex maps, we achieve exponentially fast convergence. The numerical simulations validate the feasibility of the introduced designs.

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: Space exploration embodies humanity's driver to transcend known boundaries and catalyses technological innovation, scientific advancements, and economic growth. As missions become increasingly complex, control theory emerges as a fundamental component, enhancing spacecraft navigation, operation, and adaptability in the dynamic space environment. This tutorial paper explores the pivotal role of control theory in advancing space exploration beyond the 2030s, emphasizing its contributions to autonomous decision-making, artificial intelligence, robust and resilient control, and the management of distributed systems. This paper outlines how advancements in control technologies will significantly enhance mission success and expand humanity's presence in the solar system from both fundamental research and commercial viewpoints, with contributions from academia, space agencies, and industry. This paper presents a comprehensive overview on the traditional and intelligent technological solutions, addressing current limitations in space exploration while devising future possibilities.

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: The guidance and control of space flight vehicles is undergoing a period of major transformation. With the increased emphasis on rapid low-cost access to space, and the wide array of solar system objects which are being explored by uncrewed vehicles, the need for advancements in vehicle guidance, control and performance are substantial. In this presentation, we discuss the disparate space operating environments (i.e., atmospheric, ascent, in-space, entry descent and landing (EDL), and surface exploration) and the common thread of increased autonomy that unites them. Autonomy requires building enough trust in perception, trajectory generation, and control algorithms to execute them online instead of relying on extensive offline computation and human oversight. The current barriers to autonomy vary substantially across the various space operating environments but include computational capability, “in-time” dynamically feasible trajectory generation, sufficient perception algorithms, control for perception, vehicle to vehicle teaming, and an inability to account for system and environment uncertainty using adaptive/robust/learning algorithms. The recent success of the Mars rover Perseverance and helicopter Ingenuity provide an example into the possibilities that autonomy could provide. While Ingenuity was by any measure a huge success, its most important contribution was operating as a scout for Perseverance’s next science mission. These missions required substantial human analysis and planning (over many sols). It is easy to imagine the research potential that Ingenuity could have provided if it was able to autonomously land on and to inductively charge off Perseverance’s nuclear MMRTG power supply in a certain level of autonomy was available. In this work, we make the case for investing in tools that foster collaboration between government, academia, and industry to facilitate the unification of perception, trajectory generation and control for the wide array of vehicles in the various space operating environments. Moreover, we seek to provide inroads to other researchers outside the traditional aerospace fields, to facilitate their contributions necessary to achieve the autonomy advances required by mode

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: The talk will have its roots in the field of Guidance, Navigation and Control (GNC), which is where the main developments and need for automatic control has been and applied for space systems since the beginning of space flight. It will look to the future bringing forward proven heritage control designs and expanding with new domains of real time optimisation and autonomy. Initially branching out from RendezVous and Docking (RVD) GNC systems, which are relatively mature today for routine flights, there are benefits with further maturation. This is identified to be in the direction of real time guidance optimisation enhancing adaptability as well as the benefits from a full scale multivariable control system. Relative navigation and in particular the image processing diversification with ever better soft and hardware available to better handle the difficult illumination environment in space. Today's failure handling on spacecraft has changed little over time and is mostly implemented by redundancy, but new promising Fault Tolerant Control (FTC) partly based on Robust control techniques could bring benefits and has already been tried in the aircraft domain with success at small scale. The new domains of in orbit assembly and debris removal are both leaning heavily on proven RVD control and technologies but there are needs for new developments in the directions of distributed FTC control systems, adaptive real time (re)planning. For the assembly of larger space structures developments towards an autonomy level 4, which includes decision making capability, are enabling technologies in development. Further to the above types of systems the application of multi objective optimisation techniques at interdisciplinary system level is promising. This will try to integrate, e.g., the GNC design with a part of the structures design subject to various constraints finding a Pareto based best compromise from both worlds. This can be applied to control, thruster layout, thermal issues from plumes etc. boundaries driven by several requirements. Preliminary activities have been performed paving the way further towards more complex real scale systems. Finally we will look into the benefits of advanced systems and control in some

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: The history of JAXA's rendezvous and docking missions began with the automatic docking demonstration of the ETS-VII, developed by NASDA (now JAXA). This technology was later adopted by the HTV, a cargo vehicle developed by JAXA for the ISS, which successfully carried out nine rendezvous and berthing operations. The HTV-X, the successor to the HTV, is currently under development, with an automated docking demonstration planned for HTV-X2. The results of this demonstration are expected to contribute to the future Artemis program, particularly in the context of automated docking with the Lunar Gateway. Additionally, these technological advancements hold promise for future space debris removal missions and sample return missions. This paper provides an overview of JAXA's past, present, and future rendezvous and docking missions, and discusses the related research in control engineering and machine learning. Specific topics include low-thrust autonomous rendezvous and parameter tuning methods, machine learning-based on-board navigation, and safety analysis techniques for off-nominal scenarios aimed at enhancing fault-tolerant designs.

Keywords: Aerospace, Control applications, Autonomous systems
Abstract: The expansion of space ventures highlights the need for innovative In-Orbit Servicing (IOS) and Assembly (IOA) solutions. Traditional approaches rely on rigid algorithms or expensive sensors, but these fall short in dynamic mission conditions. We propose a transformative approach integrating Artificial Intelligence (AI) into Guidance, Navigation, and Control (GNC) systems for IOS and IOA scenarios. AI-driven automation enables autonomous adaptation to changing conditions, reducing reliance on ground control and human intervention. Moreover, AI and ML are revolutionizing space missions by enhancing efficiency and autonomy. They enable real-time data analysis, optimizing navigation and decision-making processes with reduced human intervention. Autonomous systems powered by AI reduce the need for constant ground control, allowing spacecraft to adapt to unexpected situations and perform complex tasks. These technologies also improve resource management and predictive maintenance, ensuring mission effectiveness and longevity. Furthermore, space exploration based on AI will unlock new scientific discoveries and the creation of new advanced space infrastructures.