Keywords: Reinforcement learning
Abstract: Recent advances in machine learning and cooperative control have strengthened multi-UAV decision-making and coordination, enabling faster and more efficient exploration of complex, hazardous, or human-unfriendly environments. Multi-UAV cooperative target search is a critical task in scenarios such as surveillance and disaster response, yet most existing studies overlook onboard energy constraints, limiting their real-world deployment capabilities. We address this gap by designing a partially observable, energy-aware environment. In this environment, UAVs have heterogeneous initial energy distributions, and distinct energy costs for actions such as movement and hovering. Building on multi-agent deep reinforcement learning (MADRL), we propose EN-QMIX, an energy-aware extension of QMIX that integrates energy information into decision-making and designs reward mechanisms to balance coverage, target discovery, and efficiency. We further propose two additional evaluation metrics: an energy-dispersion measure and an energy-efficiency metric. Experiments show that EN-QMIX achieves more consistent energy usage, higher coverage and discoveries, and comparable or lower collisions than existing studies and baselines without considering energy constraints.

Keywords: Control applications, Agents-based systems, Data storage systems
Abstract: This paper introduces a robust, multi-objective deep reinforcement learning (DRL) controller to minimize the total operational costs of data center cooling plants. By leveraging a Dueling Double Deep Q-Network (D3QN) agent trained with domain randomization, our approach dynamically balances energy consumption, water usage, and equipment wear. Evaluated in the AlphaDataCenterCooling simulation environment, the proposed controller reduces total annual operating costs by 9.6% compared to a rule-based baseline, effectively doubling the savings of a standard energy-only DRL approach. These results highlight the necessity of multiobjective optimization and robust sim-to-real training for practical, cost-effective facility management.

Keywords: Modeling, Estimation, Simulation
Abstract: We propose a mixed-integer linear programming (MILP) framework to reduce police use-of-force incidents, designed to complement existing prevention tools such as early intervention systems (EIS). We model use-of-force incidents using a multivariate Hawkes process and optimize officer pairing networks through MILP to minimize the expected number of use-of-force events. In addition, we introduce a new risk metric, the reproduction number theta, which quantifies how likely an officer’s use-of-force incident is to trigger subsequent incidents, either by the same officer or by colleagues nearby in the network. Our results indicate that reducing use-of-force requires pairing high-theta officers with low-theta officers. By preventing clusters of high-risk officers, the optimized network suppresses social contagion of risky behaviors, disrupts escalation chains, and reduces the forecasted frequency of use-of-force incidents.

Keywords: Control applications, Data storage systems, Mechatronics
Abstract: This paper proposes a novel methodology for the design of track-following controllers in tape drive systems that operate at different tape velocities. The proposed method leads to the direct minimization of the worst-case position error signal (PES) standard deviation across different tape velocities, while properly addressing robustness considerations using a mixed H2/H∞ framework. The plant model is obtained by an exact discretization scheme and a discrete-time Linear Parameter-Varying (LPV) controller is designed. Experimental results are presented to demonstrate the effectiveness of the proposed method.

Keywords: Control applications, Hierarchical control, Data storage systems
Abstract: Optimizing data center cooling systems is challenging due to the complex hybrid action space of discrete equipment states and continuous operational setpoints. This paper proposes a novel Cooperative Hierarchical Reinforcement Learning (Co-HRL) framework to address this complexity by decomposing the problem into a two-level hierarchy. Our approach introduces a cross-entropy-based stability-aware exploration strategy for the high-level meta-controller, and a Correlated Equilibrium mechanism for cooperative low-level sub-controllers. Evaluated in the high-fidelity AlphaDataCenterCooling environment, the Co-HRL agent achieves a 16% improvement in overall energy efficiency compared to a conventional HRL baseline, demonstrating its effectiveness for complex industrial control.

Keywords: Control applications, Hybrid systems, Stability of nonlinear systems
Abstract: Planar juggling of a devil-stick using impulsive inputs is addressed using the concept of discrete virtual holonomic constraints (DVHC). The location of the center-of-mass of the devil-stick is specified in terms of its orientation at the discrete instants when impulsive control inputs are applied. The discrete zero dynamics (DZD) resulting from the choice of DVHC provides conditions for stable juggling. A control design that enforces the DVHC and an orbit stabilizing controller are presented. The approach is validated in simulation.

Keywords: Control applications, Manufacturing systems, Numerical algorithms
Abstract: This article presents a highly efficient optimal control algorithm and policies for lyophilization (aka vacuum freeze drying). The optimal solutions and control policies are derived using an extended version of simulation-based algorithm that reformulates an optimal control problem as a hybrid discrete-continuous system of mixed-index differential-algebraic equations and subsequently calculates the optimal control vector via simulation of the resulting DAEs. Our algorithm and control policies are demonstrated in a number of case studies that encompass various lyophilization and optimal control strategies. All the case studies can be solved within less than a second on a normal laptop, regardless of their complexity. The method is several orders of magnitude faster than the traditional optimization-based techniques while giving similar/better accuracy. The proposed algorithm offers an efficient and reliable framework for optimal control of lyophilization, which can be extended to similar systems with phase transitions.

Keywords: Emerging control applications, Computational methods, Modeling
Abstract: This work proposes a bid-shading strategy for first-price auctions as a measure-valued optimization problem. We consider a standard parametric form for bid shading and formulate the problem as convex optimization over the joint distribution of shading parameters. After each auction, the shading parameter distribution is adapted via a regularized Wasserstein-proximal update with a data-driven energy functional. This energy functional is conditional on the context, i.e., on publisher/user attributes such as domain, ad slot type, device, or location. The proposed algorithm encourages the bid distribution to place more weight on values with higher expected surplus, i.e., where the win probability and the value gap are both large. We show that the resulting measure-valued convex optimization problem admits a closed form solution. A numerical example illustrates the proposed method.

Keywords: Mechatronics

Keywords: Data storage systems, Robust control, Mechatronics
Abstract: In hard disk drives (HDDs), precise magnetic-head positioning is accomplished using a dual-stage actuator (DSA). These actuators are generally managed through a decoupled control structure, which makes the controller development process more straightforward. However, maintaining robust performance under system disturbances while keeping this structure intact remains a major challenge. This paper presents an alternative control design approach to resolve this issue. By applying a robust loop-shaping technique, specifically, the Robust Controller Bode (RCBode) plot, within the realm of classical control theory, the method enables engineers to design high-performance robust controllers by shaping filters directly on a Bode diagram, following clear design rules. Simulation studies based on the HDD benchmark problem confirm the effectiveness of the proposed approach, showing strong robustness against disturbances typically encountered in data center environments.

Keywords: Control education, Modeling, PID control
Abstract: The complexity of modern control systems necessitates engineering students develop a comprehensive understanding of both theoretical and practical aspects of control theory. Traditional control system education often emphasizes mathematical foundations, which can be abstract and difficult to fully grasp without hands-on experience. To bridge this gap, educational tools that combine theory with real-world applications are essential. Our prior work presented the mechanical design and preliminary validation of a novel five-bar mechanisms with variable topology for industrial automation applications. Building on that foundation, this paper develops the system modeling and control design of the mechanism and demonstrates its value as an educational tool for control systems education. The platform offers opportunities to explore non-linear dynamics, variable mechanism topologies, and hardware-in-the-loop simulations using a real-world mechanism that integrates two widely utilized configurations: a classic four-bar mechanism and a slider-crank mechanism. This combination allows students to engage with both fundamental and advanced control concepts in a practical, industry-relevant context. A pick-and-place application demonstrates the platform’s capabilities, employing a holistic approach that includes system parameter identification, the development of a supervisory state machine, and low-level control to implement a feedforward-feedback controller. Using the five-bar mechanism with variable topology as an educational tool has the potential to improve student engagement and understanding, bridging the gap between theoretical coursework and the demands of modern industrial control systems.

Keywords: Finance, Computational methods, Optimization
Abstract: This paper proposes a simulation-based framework for assessing and improving the performance of a pension fund management scheme. This framework is modular and allows the definition of customized performance metrics that are used to assess and iteratively improve asset and liability management policies. We illustrate our framework with a simple example that showcases the power of including adaptable features. We show that it is possible to dissipate longevity and volatility risks by permitting adaptability in asset allocation and payout levels. The numerical results show that by including a small amount of flexibility, there can be a substantial reduction in the cost to run the pension plan as well as a substantial decrease in the probability of defaulting.

Keywords: Finance, Stochastic optimal control, Modeling
Abstract: In this paper, we study the optimal pairs trading problem within a continuous-time mean–variance (MV) framework. The spread between two risky assets is modeled as an Ornstein–Uhlenbeck (OU) process. To mitigate potential losses arising from a non-converging spread, we incorporate a stop-loss mechanism. Owing to the time inconsistency of the mean–variance criterion, we derive an explicit equilibrium strategy by solving the extended Hamilton-Jacobi-Bellman (HJB) equation. Furthermore, we conduct numerical experiments to illustrate the equilibrium strategy and assess its performance.

Keywords: PID control, Robotics
Abstract: This work investigates optimal proportional-integral-derivative (PID) gains (Kp, Ki, Kd) to improve accuracy and precision in a six-degree-of-freedom (6-DOF) collaborative robotic arm under varying operating conditions. Experiments were conducted on the Fairino FR3 under displacements of 200-600 mm, speeds of 10-40 mm/s, and payloads of 0 and 2.4 kg. Baseline results without PID showed that steady-state error was dominated by joint compliance and backlash, with payload inertia providing slight improvements at short ranges. A stepwise empirical tuning procedure was applied at the most demanding operating points (medium-40 mm/s and far-40 mm/s) to obtain optimal gain sets. Cross-testing demonstrated that these gains generalized across the workspace, reducing steady-state error to below 0.002 mm. Results further showed that while PID control improved overall performance, simpler controllers (P, PI, PD) occasionally outperformed full PID, emphasizing the need for scenario-dependent controller selection. The study contributes an experimental framework for evaluating precision control in collaborative manipulators, a gain-scheduling strategy based on extreme-condition tuning to improve robustness, and evidence that simpler controllers can outperform full PID under certain conditions.

Keywords: Robust control, Optimization algorithms, PID control
Abstract: Magnetic levitation systems are widely recognized for their superior energy efficiency and negligible mechanical friction, making them highly suitable for modern industrial applications. Nevertheless, control of such systems poses significant challenges due to their inherently nonlinear dynamics and open-loop instability resulting from complex magnetic force interactions. This study presents a novel tuning approach for a fractional-order PI–PD controller aimed at enhancing closed-loop performance. All stability region is mapped in the (mathbb{K}_{p_{PD}}, mathbb{K}_{d_{PD}})-plane for a fixed fractional order mu, which is selected to simultaneously satisfy frequency-domain specifications, including the phase margin and gain and phase crossover frequencies. All stability region is plotted and subsequently employed as a constraint in the optimization process. Controller gains are optimized by using crayfish optimization, which guarantees system stability by restricting the gain values to lie within the predefined stability boundaries. On similar lines, outer-loop PI controller parameters are determined. Effectiveness of the proposed control scheme is validated through simulation-based performance comparisons with a recent state-of-the-art method. In addition, real-time experimental results confirm the controller’s practical feasibility and suitability for implementation in real-world scenarios.

Keywords: Identification, Adaptive control, Nonlinear systems identification
Abstract: This paper presents an online approximation framework for fully-observable nonlinear continuous-time systems using Chebyshev pseudospectral method and Chebyshev nodes. While pseudospectral methods are often applied to fixed intervals, the proposed framework is adapted over a periodic moving time window for online state approximation. In particular, the Fast Chebyshev Transform (FCT), known for its efficient computational advantage over traditional least-squares (LS) methods, is extended to the proposed moving-window framework to compute approximation coefficients in an online fashion. We provide analytical error bounds for the window-wise state approximation. Numerical results on a two-dimensional nonlinear system illustrate the performance of the proposed scheme and provide a comparison with a barycentric interpolation (BI) method adapted to online setting.

Keywords: Identification, Building and facility automation, Grey-box modeling
Abstract: Heating and cooling account for a large share of building energy use, motivating the development of advanced control strategies to optimize the operation of heating, ventilation, and air conditioning systems while maintaining occupant comfort. Model predictive control (MPC) is a promising approach, but it relies on a thermal model of the building. An accurate thermal model can be challenging to obtain due to unmeasured, time-varying disturbances, such as occupant activity and appliance use. This work proposes a three-step methodology. First, parameters of a thermal resistance-capacitance (RC) building model and an initial disturbance model are estimated from data. Second, an extended data set collected during regular operation is used to train a predictive disturbance model that captures daily patterns. Third, the RC state-space model is augmented with a disturbance state that dynamically corrects forecast errors using real-time measurement mismatches, yielding an integrated estimator-predictor framework consistent with its intended use in MPC. The proposed methodology is evaluated using data from a multi-family residential unit equipped with a variable-speed heat pump, demonstrating the prediction accuracy of the resulting estimator-predictor modeling framework.

Keywords: Identification, Estimation, Stochastic systems
Abstract: Hawkes processes with user-specified unit-mass causal kernels admit a linear parametrization that supports least-squares (LS) identification. A closed-form LS estimator exists only when the empirical Gram matrix is positive-definite. However, this is only guaranteed with high probability. Here, we study a continuous-time LS, and prove the Gram is positive definite almost surely for finite samples, yielding closed-form estimators. Under standard regularity, the estimators are shown to be consistent under correct specification, and converge to well-defined pseudo-true parameters in the general case.

Keywords: Identification, Hybrid systems, Optimization
Abstract: This paper investigates the problem of identifying state-dependent switching systems, a class of hybrid dynamical systems that combine multiple linear or nonlinear modes. We propose two broad classes of switching systems: switching linear systems (SLSs) and switching polynomial systems (SPSs). We first formulate the joint estimation of the mode dynamics and switching rules as a mixed integer program. To solve its inherent scalability issue, we develop a hierarchy of convex relaxations and establish a bound and conditions under which these relaxations are tight. Building on these results, we propose a bilevel convex optimization framework that alternates between mode assignment and dynamics estimation, and we recover switching boundaries using margin-based polynomial classifiers. Numerical experiments on both linear and nonlinear oscillators demonstrate that the method accurately identifies mode dynamics and reconstructs switching surfaces from trajectory data. Our results provide a tractable optimization-based framework for switching system identification.

Keywords: Identification, Identification for control
Abstract: Forgetting is an essential component of learning. This paper provides a numerical comparison of several variable-rate forgetting (VRF) methods within the context of recursive least squares (RLS) for online system identification. A secondary goal of the paper is to compare the relative effectiveness of the quadratic residual and weighted quadratic residual. These residuals are used with five forgetting methods, namely, single-point residual (SPR), single-window regression (SWR), dual-window RMS (DWR), dual-window variance (DWV), and adjacent-window RMS (AWR), where SWR, DWR, DWV, and AWR take advantage of kick-on, kick-off logic.

Keywords: Identification, Linear systems, Numerical algorithms
Abstract: In practical least squares realization problems, partial information about the pole locations of the dynamical model may be known a priori. Existing techniques for incorporating this prior knowledge, such as prefiltering the given data, are typically heuristic and lack theoretical guarantees. We extend our previously developed globally optimal estimation approach to accommodate fixed poles in the least squares realization problem. In particular, we reformulate the problem as a (rectangular) multiparameter eigenvalue problem, the eigenvalues of which characterize all local and global minimizers of the constrained estimation problem. We present numerical examples to demonstrate the effectiveness of the proposed method and experimentally validate the paper's central hypothesis: incorporating a priori information on the poles enhances the estimation results.

Keywords: Identification for control, Networked control systems, Autonomous robots
Abstract: Origami structures exhibit complex dynamical behaviors during reconfiguration due to coupled panel deformation, hinge rotations, and folding-induced changes in stiffness and compliance. This work addresses origami reconfiguration control by developing an integrated control and model identification framework. First, an extended first-order generalized pseudo-Bayesian (E-GPB1) filter is developed to jointly estimate continuous states and identify discrete dynamic modes. Second, a probability-weighted state-dependent Riccati equation (SDRE) controller is designed to regulate the system under input constraints. The combined E-GPB1/SDRE framework enables robust control despite switching dynamics and physical parameter variations. %Theoretical analysis of the estimation error and closed-loop stability convergence is provided. In addition, simulation on a Kresling origami pattern demonstrates accurate mode identification, reliable trajectory regulation, and stable convergence to an otherwise unstable configuration. These results establish the proposed approach as an effective strategy for estimation and control of reconfigurable origami-inspired structures.

Keywords: Identification for control, Quantized systems, Robust control
Abstract: Cloud-assisted system identification and control have emerged as practical solutions for low-power, resource-constrained control systems such as micro-UAVs. In a typical cloud-assisted setting, state and input data are transmitted from local agents to a central computer over low-bandwidth wireless links, leading to quantization. This paper investigates the impact of state and input data quantization on a linear time invariant (LTI) system identification, derives a worst-case bound on the identification error, and develops a robust controller for guaranteed cost control. We establish a fundamental bound on the model error that depends only on the quantized data and quantization resolution, and develop a linear matrix inequality (LMI) based guaranteed cost robust controller under this error bound.

Keywords: Differential-algebraic systems, Nonlinear systems identification, Observers for nonlinear systems
Abstract: In this work, we extend the sensitivity-based rank condition (SERC) test for local observability to an- other class of systems, namely smooth and nonsmooth differential-algebraic equation (DAE) systems of index-1. The newly introduced test for DAEs, which we call the lexicographic SERC (L-SERC) observability test, utilizes the theory of lexicographic differentiation to compute sensitivity information. Moreover, the newly introduced L-SERC observability test can judge which states are observable and which are not. Additionally, we introduce a novel sensitivity- based extended Kalman filter (S-EKF) algorithm for state estimation, applicable to both smooth and nonsmooth DAE systems. Finally, we apply the newly developed S-EKF to estimate the states of a wind turbine power system model.

Keywords: Nonlinear systems identification, Fluid flow systems, Reduced order modeling
Abstract: Measuring sediment transport in riverbeds has long been a challenging research problem in geomorphology and river engineering. Traditional approaches rely on direct measurements using sediment samplers. Although such measurements are often considered ground truth, they are intrusive, labor-intensive, and prone to large variability. As an alternative, sediment flux can be inferred indirectly from the kinematics of migrating bedforms and temporal changes in bathymetry. While such approaches are helpful, bedform dynamics are nonlinear and multiscale, making it difficult to determine the contributions of different scales to the overall sediment flux. Fourier decomposition has been applied to examine bedform scaling, but it treats spatial and temporal variability separately. In this work, we introduce Dynamic Mode Decomposition (DMD) as a data-driven framework for analyzing riverbed evolution. By incorporating this representation into the Exner equation, we establish a link between modal dynamics and net sediment flux. This formulation provides a surrogate measure for scale-dependent sediment transport, enabling new insights into multiscale bedform-driven sediment flux in fluvial channels.

Keywords: Nonlinear systems identification, Identification, Identification for control
Abstract: We consider federated learning of linearly- parameterized nonlinear systems. We establish theoretical guarantees on the effectiveness of federated nonlinear system identification compared to centralized approaches, demonstrating that the convergence rate improves as the number of clients increases. Although the convergence rates in the linear and nonlinear cases differ only by a constant, this constant depends on the feature map ϕ, which can be carefully chosen in the non- linear setting to increase excitation and improve performance. We experimentally validate our theory in physical settings where client devices are driven by i.i.d. control inputs and control policies exhibiting i.i.d. random perturbations, ensuring non-active exploration. Experiments use trajectories from non- linear dynamical systems characterized by real-analytic feature functions, including polynomial and trigonometric components, representative of physical systems including pendulum and quadrotor dynamics. We analyze the convergence behavior of the proposed method under varying noise levels and data distributions. Results show that federated learning consistently improves convergence of any individual client as the number of participating clients increases.

Keywords: Nonlinear systems identification, Lyapunov methods, Adaptive control
Abstract: This paper presents a concurrent learning-based, high-order tuner adaptive control framework for the end effector tracking control of robotic manipulators. Via defining the tracking error in task space, the proposed controller circumvents the computational intensity of resolving position level inverse kinematic solutions at each sampling instance. A Lyapunov-based stability analysis verifies the exponential stability of both parameter estimation and task space tracking errors. The effectiveness of the controller and adaptation algorithm is validated through numerical studies conducted on the model of a two link, revolute joint robotic arm, with a focus on trajectory tracking and parameter identification performance.

Keywords: Nonlinear systems identification, Machine learning, Numerical algorithms
Abstract: This manuscript presents a framework for resolving inverse problems through the use of operator approximations over vector valued RKHSs. This generalizes Koopman based methods for data driven analysis and identification of dynamical systems. Three examples of this framework are presented to highlight its generality and effectiveness.

Keywords: Nonlinear systems identification, Machine learning, Robotics
Abstract: This paper presents a control framework for input-affine fractional-order systems with unknown, nonlocal nonlinearities. We leverage Kolmogorov–Arnold Networks (KANs) to model these complex dynamics. Unlike traditional networks, the learnable, spline-based activation functions on KAN edges provide the expressive power needed to accurately capture the hereditary properties of fractional systems. The core theoretical contribution is a rigorous stability analysis of the closed-loop system, for which we establish sufficient conditions that guarantee the existence, uniqueness, and uniform stability of solutions. This analysis is validated through numerical simulations, and the framework's practical effectiveness is demonstrated on a mobile robot path-tracking problem, where a KAN-based controller successfully compensates for complex slip effects to significantly improve tracking accuracy.

Keywords: Nonlinear systems identification, Modeling, Filtering
Abstract: Noise contamination in input and state measurements complicates the data-driven identification of nonlinear systems for control and prediction. This work proposes an extension of a data-driven Koopman modeling approach, specifically to controlled nonlinear systems with unknown noise present. The proposed method uses block coordinate descent to solve a regularized least-squares problem, jointly learning a bilinear lifted-state representation and identifying the underlying noise components. This framework, analogous to eDMD but combined with a log-likelihood characterization of noise, identifies system matrices for accurate state prediction. The efficacy of the model identification technique is demonstrated on a continuously stirred tank reactor case study, achieving more than a twofold improvement in prediction accuracy over the existing extension of eDMD to controlled systems without accounting for noise characteristics.

Keywords: Nonlinear systems identification, Networked control systems, Neural networks
Abstract: In many ecological, biological, and social networks, the system dynamics is shaped not only by pairwise couplings but also by higher-order interactions which involve multiple agents simultaneously. However, identifying such higher-order interactions from time-series data remains a significant challenge. In this letter, we propose a novel framework for identifying hypergraph structures using hypergraph physics-informed neural network (HyperPINN). We model the hypergraph dynamics using continuous-time nonlinear differential equations that incorporate both pairwise and higher-order terms. By embedding the governing dynamics into the loss function of neural networks, HyperPINN jointly estimates the state trajectories and infers the unknown hypergraph structure directly from observed time-series data. Numerical experiments on coupled Rössler oscillators and Lotka-Volterra dynamics demonstrate that HyperPINN reliably uncovers hypergraph structures even under noise and limited data, outperforming state-of-the-art methods.

Keywords: Nonlinear systems identification, Statistical learning, Estimation
Abstract: We tackle the problem of system identification, where we select inputs, observe the corresponding outputs from the true system, and optimize the parameters of our model to best fit the data. We propose a practical and computationally tractable methodology that is compatible with any system and parametric family of models. Our approach only requires input-output data from the system and first-order information of the model with respect to the parameters. Our approach consists of two modules. First, we formulate the problem of system identification from a Bayesian perspective and use a linear Gaussian model approximation to iteratively optimize the model’s parameters. In each iteration, we propose to use the input-output data to tune the covariance of the linear Gaussian model. This online covariance calibration stabilizes fitting and signals model inaccuracy. Secondly, we define a Gaussian-based uncertainty measure for the model parameters, which we can then minimize with respect to the next selected input. We test our method with linear and nonlinear dynamics.

Keywords: Nonlinear systems identification
Abstract: The effectiveness of data-driven techniques heavily depends on the input signal used to generate the estimation data. However, a significant research gap exists in the field of input design for nonlinear dynamic system identification. In particular, existing methods largely overlook the minimization of the generalization error, i.e., model inaccuracies in regions not covered by the estimation dataset.

This work addresses this gap by proposing an input design method that embeds a novel optimality criterion within a receding horizon control (RHC)-based optimization framework. The distance-based optimality criterion induces a space-filling design within a user-defined region of interest in a surrogate model's input space, requiring only minimal prior knowledge. Additionally, the method is applicable both online, where model parameters are continuously updated based on process observations, and offline, where a fixed model is employed.

The space-filling performance of the proposed strategy is evaluated on an artificial example and compared to state-of-the-art methods, demonstrating superior efficiency in exploring process operating regions.

Keywords: Machine learning, Stability of nonlinear systems, Power electronics
Abstract: Learning the governing equations of dynamical systems from data has drawn significant attention across diverse fields, including physics, engineering, robotics and control, economics, climate science, and healthcare. Sparse regression techniques, exemplified by the Automatic Regression for Governing Equations (ARGOS) framework, have demonstrated effectiveness in extracting parsimonious models from time series data. However, real-world dynamical systems are driven by input control, external forces, or human interventions, which standard ARGOS does not accommodate. To address this, we introduce ARGOS with control (ARGOSc), an extension of ARGOS that incorporates external control inputs into the system identification process. ARGOSc extends the sparse regression framework to infer governing equations while accounting for the effects of exogenous inputs, enabling robust identification of forcing dynamics in low- to medium-noise datasets. We demonstrate ARGOSc’s efficacy on benchmark systems, including the Van der Pol oscillator, Lotka-Volterra, and the Lorenz system with forcing and feedback control, showing enhanced accuracy in discovering governing laws. Under the noisy conditions, ARGOSc outperforms the widely used sparse identification of nonlinear dynamics with control (SINDYc), in accurately identifying the underlying forced dynamics. In some cases, SINDYc fails to capture the true system dynamics, whereas ARGOSc consistently succeeds.

Keywords: Optimal control, Control software, Robotics
Abstract: We present COPP, an open-source, ultra-fast, and "plug-and-play" Rust library for convex-objective path parameterization (COPP). The library supports general objectives and constraints for both second- and third-order COPP. A novel method, ARDP, is proposed to solve the general COPP problem efficiently, making it suitable for online application. Specifically, COPP can solve global optimal solutions for convex problems and KKT stationary solutions for non-convex problems.

Keywords: Energy systems, Model Validation, Modeling
Abstract: This paper presents the design and modeling of a phase change material (PCM) sleeve enclosed in a Polyethylene Terephthalate Glycol (PETG) case for thermal management of lithium-ion batteries. A five-state nonlinear model captures conduction, latent heat absorption, and convective heat loss while preserving energy consistency. Experimental validation using a 26650 Li-ion cell cycled at 11 A shows strong agreement, with RMSE of 0.56◦C at the PETG inner wall and 1.44◦C in the PCM. The sleeve reduces temperature rise by ∼ 5◦C and attenuates oscillations, demonstrating effective passive thermal buffering.

Keywords: Biomedical, Feedback linearization, Reduced order modeling
Abstract: This poster presents a reduced-order dynamic modeling and control framework for three-dimensional human eye movements. Human ocular motion obeys geometric constraints such as Listing’s Law, which couples torsion with horizontal and vertical gaze. Integrating these geometric constraints with physical dynamics remains challenging.

In this work, we derive a torque-driven rigid-body formulation of the eyeball and embed Listing’s constraint directly into the system dynamics. The original six-state rigid-body model is thereby reduced to a four-state nonlinear system describing yaw and pitch dynamics, while torsion becomes a dependent variable determined by gaze orientation.

A nonlinear state-feedback controller based on input–output feedback linearization is then designed to regulate the resulting system. Controller gains are identified from human saccade trajectories using a second-order dynamic model. Simulation results demonstrate physiologically consistent three-dimensional eye movements that satisfy the geometric constraints of ocular kinematics.

Keywords: Agents-based systems, Sensor fusion, Aerospace
Abstract: This work presents a magnetic anomaly-based localization method for UAVs operating in GNSS-denied environments using a collaborative multi-agent approach that integrates magnetic and inter-UAV distance measurements within a Particle Filter framework via Bayesian inference. The method leverages inter-UAV ranging through ad hoc communications, making it well-suited for dynamic formations and resilient to communication disruptions and UAV failures. Following an introduction to the multiagent mission problem and a formal description of the proposed Distributed Particle Filter algorithm, a case study is presented using a magnetic field map and flight logs collected experimentally with a dedicated VTOL platform equipped with a high-precision scalar magnetometer operating over the Baltic Sea. Evaluation of various collaborative scenarios demonstrates that incorporating inter-UAV ranging significantly improves localization accuracy, reducing positioning error by 71-78% across different collaboration strategies compared to the non-collaborative baseline.

Keywords: Adaptive control
Abstract: False data injection (FDI) attacks present a serious threat to autonomous multi-agent systems (MASs). Existing attack-resilient control approaches typically depend on restrictive assumptions regarding attack signals and often neglect critical physical safety requirements, such as collision avoidance. In addition, conventional controllers based on Control Lyapunov Functions (CLFs) and Control Barrier Functions (CBFs) tend to compromise stability by introducing slack variables when safety and stability objectives are in conflict.

To address these limitations, this paper proposes an integrated safety-aware and attack-resilient (SAAR) control framework for heterogeneous MASs subject to exponentially unbounded false data injection (EU-FDI) attacks. The proposed approach incorporates real-time online adaptation in both the observer layer (OL) and the control input layers (CIL), enabling simultaneous mitigation of two sources of performance degradation: the detrimental impact of EU-FDI attacks and the bounded mismatch arising from safety-enforcing control inputs.

By explicitly accounting for safety-induced input deviations and attack compensation errors within a rigorous Lyapunov-based stability analysis, it is shown that the closed-loop system achieves uniformly ultimately bounded (UUB) containment tracking. Furthermore, the enforcement of CBF-based safety constraints does not undermine the stability objective, thereby eliminating the need for slack variables.

Finally, the effectiveness of the proposed framework is demonstrated through comprehensive 3D simulations, representing, to the best of our knowledge, the first study to simultaneously ensure guaranteed physical safety and resilience under concurrent EU-FDI attacks on both the OL and CIL.

Keywords: Predictive control for linear systems, Optimal control, Optimization algorithms
Abstract: The alternating direction method of multipliers (ADMM) has gained increasing popularity in embedded model predictive control (MPC) due to its code simplicity and pain-free parameter selection. However, existing ADMM solvers either target general quadratic programming (QP) problems or exploit sparse MPC formulations via Riccati recursions, which are inherently sequential and therefore difficult to parallelize for long prediction horizons. This poster proposes a novel parallel-in-horizon and construction-free nonlinear MPC algorithm, termed piMPC, which combines a new variable-splitting scheme with a velocity-based system representation in the ADMM framework, enabling horizon-wise parallel execution while operating directly on system matrices without explicit MPC-to-QP construction. Numerical experiments and accompanying code are provided to validate the effectiveness of the proposed method.

Keywords: Distributed control, Power systems, Adaptive control
Abstract: This poster proposes a fully distributed, privacy-preserving, and safety-aware attack-resilient (P2SA2R) control framework for islanded inverter-based AC microgrids. The method ensures frequency regulation, voltage containment, and power sharing while withstanding a broad class of cyberattacks, including exponentially unbounded false data injection (EU-FDI) attacks on control inputs. Unlike prior work that addresses either transient safety or steady-state resilience in isolation, the proposed framework guarantees both, maintaining system states ultimately within safety bounds throughout the attacks. This is particularly critical for low-inertia inverter-based microgrids, which are highly susceptible to large overshoots and severe fluctuations during transients caused by disturbances or cyber threats. To address the vulnerability of information leakage due to eavesdropping in communication networks, a dynamic attack-aware output masking mechanism is integrated into the secondary control layer. This approach preserves privacy during information exchange with immediate neighbors in distributed control, without compromising consensus accuracy. Leveraging the privacy-preserved, an intelligent and fully distributed attack-resilient control framework is developed. A quadratic program (QP) is formulated to achieve the multiple objectives of the proposed P2SA2R strategy. By unifying Control Barrier Functions (CBFs) and Control Lyapunov Functions (CLFs), a rigorous theoretical analysis guarantees both ultimate safety and uniformly ultimately bounded (UUB) resilience under adversarial conditions. Simulation results on a modified IEEE 34-bus system confirm the framework’s effectiveness in achieving safety, privacy, and resilience under severe cyber threats.

Keywords: Reinforcement learning, Neural networks, Machine learning
Abstract: In the study of deep learning, the manifold hypothesis is a foundational assumption suggesting that high-dimensional data lies on a low-dimensional manifold where reasoning and representation occur. However, recent research has begun to challenge this, suggesting that token embeddings may instead inhabit stratified spaces, complex geometric structures where the local dimension can vary from point to point. This paper extends that inquiry into the realm of data driven control methods, such as Deep Reinforcement Learning (DRL). While previous studies focused on text-based tokens, this work examines a transformer-based Proximal Policy Optimization model trained to play a visual coin collecting game. We utilize a modified version of the Searing Spotlights environment from the Memory Gym, dubbed the "Two-Coin" game. In this environment, an agent must navigate a room to collect coins while avoiding dynamic, health-draining spotlights. The main objective is to determine if the latent space of a vision-based DRL agent exhibits the same non-manifold, stratified characteristics found in LLMs and to interpret how these geometric properties. By applying a Volume Growth Transform (VGT) we provide numerical evidence that the latent space is neither a manifold nor a fiber bundle and via a new realization theorem we show that the signature is consistent with that of stratified spaces. The VGT measures the local dimension of the latent space and associates a log-log volume-radius growth curve to each point. Numerical results reveal that the DRL agent's embedding space is not uniform. By tracking the local dimension over trajectories, we found that the latent representation alternates between periods of low local dimension, where the agent follows a fixed sub-strategy, and periods with high local dimension occurring when the agent is close to a sub-goal (collecting a coin) or when environmental complexity increases, i.e., when the number of spotlights increases. We believe that the study of the geometric and topological structure of the latent space is critical to better characterize DRL properties and the VGT can serve as a geometric indicator of control task complexity of such models and, generally, to study stratified spaces.

Keywords: Distributed parameter systems, Stability of hybrid systems, Computational methods
Abstract: We consider two serially-connected wave equations and compare two different transmission control designs. The first is a continuous transmission design consisting of higher-order and lower-order interface damping and the second is a discontinuous design consisting only of lower-order damping. The continuous design never produces exponential stability while the discontinuous design produces unconditional exponential stability.

Keywords: Hybrid systems, Stability of hybrid systems, Distributed parameter systems
Abstract: We study a SCOLE-inspired [Taylor, Balakrishnan'87] hybrid PDE-ODE beam-mass-beam model in which two Euler-Bernoulli beams are connected through an interior interface carrying both a translational mass and a rotational inertia [Ashburn, Brown, Ozer'26]. We propose an interface-only feedback law based on mixed-order velocity traces and prove exponential decay of the total energy via a Lyapunov approach.

Keywords: Machine learning, Kalman filtering, Numerical algorithms
Abstract: A fundamental problem in Kalman filtering, neural network verification, and inference on uncertain inputs is characterizing the distribution of outputs when inputs follow a Gaussian distribution and are transformed by a smooth nonlinear function. Assumed Density Filtering approximates hidden layers by Gaussian distributions via moment matching, but the quality of this approximation remains unclear. We address this by bounding the Wasserstein distance between transformed outputs and the Gaussian distribution with matching mean and covariance for single-layer transformations.

Applying a second-order Poincaré inequality, we derive bounds on the Gaussianity of the output distribution. However, this bound is fragile when the output distribution has principal directions of low variance. At face value, the second-order Poincaré inequality is weaker than the asymptotic delta method.

Our key insight is to decompose outputs via principal component analysis on the output covariance. Projecting onto the dominant eigenspaces isolates the non-degenerate subspace, yielding a bound that separates an approximately Gaussian component from a rapidly-decaying residual.

This technique establishes that moment matching outperforms Jacobian linearization even in asymptotic regimes where both agree to leading order, and enables layerwise Wasserstein recursion guarantees for deeper networks.

Keywords: Predictive control for linear systems, Stability of linear systems, Linear systems
Abstract: This paper addresses the problem of continuous-time model predictive control (MPC). We show that in the case of linear time-invariant (LTI) systems, if the input trajectory is parametrized using B-splines and if the cost function is quadratic and if the reference trajectory is constant over the optimization horizon, then the continuous-time MPC problem can be solved off-line in closed-form, and that the resulting control strategy has a classic feedback structure. We demonstrate the method using two simple examples. The advantages of our approach are that on-line optimization is not required, the time horizon can be arbitrarily long, and fidelity of the control solution can be selected independent of the time horizon.

Keywords: Game theory, Agents-based systems, Stability of linear systems
Abstract: Multi-agent decision-making involves agents making strategic decisions while considering or predicting each other’s behavior. Model predictive games are a class of controllers in which an agent iteratively solves a finite-horizon game to predict the collective behavior of a multi-agent system and synthesize their own control action from the first time step in a receding-horizon fashion. When multiple agents implement these types of controllers, there may exist misspecifications in their respective game models resulting from not accurately knowing other agents’ objectives. This work studies the effects of these prediction misalignments by providing criteria for the stability of systems of agents deploying model predictive game controllers with heterogeneous game models.

Keywords: Nonlinear systems identification, Identification for control, Spacecraft control
Abstract: This work examines the merits of a new method of nonlinear system identification (NLSID) designed to work with extremely limited memory availability. Spacecraft commonly exhibit nonlinear flexible dynamics that must be learned or validated after launch. These spacecraft often have extremely limited onboard computational resources due to a combination of power and mass limitations and radiation hardness requirements. This confluence of requirements and limitations drives the need for a NLSID algorithm that can be run with the extremely memory-limited hardware onboard the spacecraft. The proposed algorithm was developed using tools from nonlinear systems theory and statistical stepwise regression. The learned dynamic models are built up using a forward selection process, which is then followed by a backward selection step to avoid superfluous terms in the final model. Both forward and backward selection steps use the Bayesian information criterion as a selection metric. The algorithm is named for this process, as the Iterative Additive Minimization of the Bayesian Information Criterion (IAMBIC) . The IAMBIC algorithm is demonstrated on three systems of increasing complexity and difficulty: a simple linear system, the Lorenz chaotic oscillator, and a high-order solar sail model.

Keywords: Autonomous systems, Control of networks, Adaptive control
Abstract: In multi-robot systems, maintaining persistent communication graph connectivity is often overly restrictive, especially when robots have limited communication ranges but operate in large environments. Instead, allowing robots to temporarily disconnect and later reconnect is often more desirable for efficient task execution while still ensuring timely information sharing across the team. In this paper, we propose a novel prescribed-time control barrier function framework that enables robots to temporarily disconnect and re-enter the communication range within an adjustable and feasible prescribed time. Moreover, we introduce a reconnection triggering mechanism that jointly considers task execution and reconnection urgency, thereby providing a principled way to decide when reconnection should occur. Theoretical analysis justifies convergence to the satisfying reconnection within a prescribed finite time. Experimental results validate the performance of our proposed adaptive PT-CBF with improved task efficiency and satisfying reconnections.

Keywords: Robotics, Uncertain systems, Optimal control
Abstract: This paper presents a control framework for the human oculomotor system that addresses nonlinear dynamics, inertia uncertainty, and the physiological constraints inherent in eye movement. Accurate eye control is essential for applications such as biomimetic robotic eyes and prosthetic systems. However, existing models often lack effective closed-loop control and fail to incorporate important physiological principles such as Listing’s law. Additionally, the non-uniform structure of the eye introduces uncertainty in the inertia matrix, including both diagonal and off-diagonal coupling terms, which complicates closed-loop control design. To address these challenges, a decoupled linear-nonlinear control strategy is proposed. The oculomotor dynamics are reformulated into a structured state-space model that enables the design of an optimal control law based on linear quadratic regulation principles. The approach separates the linear and nonlinear dynamics, allowing robust compensation for model uncertainties. Furthermore, a correction mechanism is introduced to enforce Listing’s law, ensuring that the resulting eye motion remains physiologically consistent while achieving the desired tracking performance. Simulation results demonstrate that the proposed controller achieves accurate trajectory tracking with negligible error while maintaining control inputs within practical limits. The method reduces the effect of inertia uncertainty and preserves the natural constraints of eye movement. These findings support the potential of the proposed framework for robotic and prosthetic eye-control applications, offering both improved tracking performance and physiological consistency.

Keywords: Robotics, Decentralized control, Autonomous robots
Abstract: Decentralized safe control plays an important role in multi-agent systems given the scalability and robustness without reliance on a central authority. However, without an explicit global coordinator, the decentralized control methods are often prone to deadlock-a state where the system reaches equilibrium, causing the robots to stall. In this paper, we propose a generalized decentralized framework that unifies the Control Lyapunov Function (CLF) and Control Barrier Function (CBF) to facilitate efficient task execution and ensure deadlock-free trajectories for the multi-agent systems. As the agents approach the deadlock-related undesirable equilibrium, the framework can detect the equilibrium and drive agents away before that happens. This is achieved by a secondary deadlock resolution design with an auxiliary CBF to prevent the multi-agent systems from converging to the undesirable equilibrium. To avoid dominating effects due to the deadlock resolution over the original task-related controllers, a deadlock indicator function using CBF-inspired risk measurement is proposed and encoded in the unified framework for the agents to adaptively determine when to activate the deadlock resolution. This allows the agents to follow their original control tasks and seamlessly unlock or deactivate deadlock resolution as necessary, effectively improving task efficiency. We demonstrate the effectiveness of the proposed method through theoretical analysis, numerical simulations, and real-world experiments.

Keywords: Feedback linearization, Biomedical, MEMs and Nano systems
Abstract: Noncontact magnetic manipulators utilize arrays of multiple magnets to generate and precisely control magnetic fields, which are then exploited to manipulate magnetic objects from a distance without direct contact. This unique noncontact feature provides a viable means for safe and precise operation of magnetically driven tools inside the human body for new generations of minimally invasive surgical, imaging, and drug targeting procedures. Prior work on design and feedback control of magnetic manipulators mostly focuses on the manipulation of single magnetic objects, while simultaneous control of multiple objects can promote a broad range of new applications. This work is aimed at the development of feedback control techniques that enable simultaneous manipulation of multiple magnetic objects. Early experimental results of this work demonstrate successful steering of two magnetic particles along desired reference trajectories. The control techniques developed for the experiments can be extended to more than two particles, provided that the magnetic manipulator in use incorporates enough number of independently controlled magnets.

Keywords: Robotics, Autonomous robots, Reinforcement learning
Abstract: Safety is a critical concern in learning-enabled autonomous systems especially when deploying these systems in real-world scenarios. An important challenge is accurately quantifying the uncertainty of unknown models to generate provably safe control policies that facilitate the gathering of informative data, thereby achieving both safe and optimal policies. Additionally, the selection of the data-driven model can significantly impact both the real-time implementation and the uncertainty quantification process. In this paper, we propose a provably sample efficient episodic safe learning framework that remains robust across various model choices with quantified uncertainty for online control tasks. Specifically, we first employ Quadrature Fourier Features (QFF) for kernel function approximation of Gaussian Processes (GPs) to enable efficient approximation of unknown dynamics. Then the Adaptive Conformal Prediction (ACP) is used to quantify the uncertainty from online observations and combined with the Control Barrier Functions (CBF) to characterize the uncertainty-aware safe control constraints under learned dynamics. Finally, an optimism-based exploration strategy is integrated with ACP-based CBFs for safe exploration and near-optimal safe nonlinear control. Theoretical proofs and simulation results are provided to demonstrate the effectiveness and efficiency of the proposed framework.

Keywords: Robotics, Learning, Robust control
Abstract: We develop a safety framework for port-Hamiltonian systems that combines graph-based barrier synthesis with Bayesian uncertainty propagation. Using only the known interconnection, dissipation, and input topology, we construct an influence graph whose shortest-path distance from constrained to actuated compartments lower-bounds the relative degree of a safety constraint, and reveals structural obstructions when no path exists. For higher-relative-degree constraints, we reshape the specification by subtracting shifted local storage terms from a barrier-insulating blanket, producing a port-transversal barrier of relative degree one and recovering CBF feasibility under suitable regularity conditions and bounded inputs. When the Hamiltonian is learned from data, we propagate uncertainty through the same structure using posterior credible bands on local energy storages and a separate drift credible set in the CBF condition, yielding safety guarantees with independently tunable credibility budgets. Experiments on LC ladder networks and learned mechanical systems demonstrate that enforceability is governed by topology, while the Bayesian extension preserves safety under model uncertainty with reduced conservatism.

Keywords: Autonomous systems, Estimation, Observers for nonlinear systems
Abstract: This paper develops a nonlinear observer for estimating the full pose and velocities of a rigid body undergoing coupled rotational and translational motion in three dimensions. The proposed method is constructed directly on the tangent bundle of the special Euclidean group, SE(3), thereby avoiding the use of local coordinates or quaternion-based representations and eliminating issues such as singularities and unwinding. The estimator relies on onboard sensor data, including 3D position vectors referenced to inertially-fixed points and angular velocity measurements expressed in the body frame. Finite-time convergence of the estimation errors is established through a Lyapunov-based analysis, ensuring rapid and robust performance in the presence of bounded measurement noise. Experimental results are obtained using point cloud data and rate gyro measurements collected from the ZED 2i stereo depth camera. The FTS-PE is compared with our previously developed variational pose estimator (VPE), and experimental results show the stability and robustness of both methods, as well as the faster convergence of the FTS-PE. The proposed framework has applications to spacecraft autonomous rendezvous and proximity operations (ARPO) and autonomous refueling of a fixed-wing unmanned aerial vehicle (UAV) with a fuel tanker, both of which require terminal docking maneuvers, in which a vehicle must estimate its pose relative to another during close proximity operations.

Keywords: Machine learning, Smart grid, Transportation networks
Abstract: The rapid integration of electric vehicles (EVs) is foundational to sustainable transportation but introduces unprecedented volatility, non-stationarity, and spatial complexity into power distribution networks. Accurate short-term forecasting of EV charging demand is critical for dynamic grid stability; however, traditional deep learning and standard spatio-temporal graph convolutional networks (STGCNs) fundamentally struggle to decouple underlying structural charging trends from high-frequency stochastic noise, frequently acting as low-pass filters that fail to capture extreme localized demand transients. To address this limitation, this study proposes a novel multi-channel forecasting framework that integrates complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) into an STGCN architecture to explicitly isolate and spatially route complex frequency dynamics. The proposed methodology processes continuous hourly EV charging loads across a network of functional charging nodes by utilizing CEEMDAN to decompose the highly volatile, raw energy signal into twelve distinct intrinsic mode functions (IMFs). This forecasting model successfully captures extreme transient load spikes while adhering to zero-bound grid constraints. Our work contributes towards a robust predictive architecture that substantially enhances the reliability of vehicle-to-grid (V2G) integration.

Keywords: Automotive control, Automotive systems, Traffic control
Abstract: Reliable motion prediction is fundamental to autonomous driving, as future traffic evolution directly determines the safety and feasibility of control decisions. Learning-based approaches, such as occupancy and flow prediction, capture rich interaction dynamics for short-term forecasting but suffer from error accumulation over longer horizons. In contrast, physics-based approaches, including queue-theoretic models, provide structured and stable long-term estimates but lack fidelity in representing short-term interactions. Existing methods typically address these regimes separately, limiting their effectiveness for control-oriented prediction. This paper proposes a hybrid virtual-leader prediction framework that unifies learning-based short-term interaction modeling with queue-theoretic long-term traffic structure to enable reliable prediction over extended horizons. The virtual leader is formulated as a synthesized trajectory representing the aggregated influence of surrounding traffic and serves as a direct acceleration reference for longitudinal control. For the short-term horizon (0–5 s), a learning-based model predicts occupancy and flow, which are incorporated into an optimization framework that balances safety (collision avoidance) and efficiency (speed tracking) to infer acceleration. For the long-term horizon (5–15 s), a queue-theoretic model estimates traffic evolution under consistent phase assumptions, generating stable acceleration references. The framework is evaluated on the Waymo Motion Dataset and demonstrates improved performance compared to queue-theory-only baselines across diverse scenarios, including normal driving, congestion, and shockwave propagation. Results show reduced error accumulation and more consistent long-horizon behavior while maintaining short-term interaction fidelity. This work provides a practical pathway to integrate data-driven prediction with physically grounded models for control-oriented autonomous driving systems.

Keywords: Optimization, Optimization algorithms, Transportation networks
Abstract: Submodular optimization provides a powerful framework for modeling diminishing-returns phenomena in machine learning, networked systems, and resource allocation. However, most existing formulations assume a single ground set, limiting their applicability to real-world cyber-physical systems where decisions naturally span multiple interacting domains. In this work, we introduce k-matroid submodularity, a novel class of set functions defined over multiple ground sets, enabling a principled treatment of joint allocation and routing decisions without resorting to computationally expensive product-space constructions. We formulate a transportation allocation problem as a k-matroid submodular maximization task, where items must be distributed across multiple providers while simultaneously selecting feasible transportation routes in a network. This formulation captures both diminishing returns in allocation utilities and increasing costs in transportation, unifying them into a single structured optimization problem. To solve this class of problems, we propose COCO-Greedy, a coordinate-wise continuous greedy algorithm that leverages the inherent hierarchy across matroids. The method operates on the multilinear extension of the objective and performs stochastic gradient updates over individual ground sets, followed by pipage rounding to recover feasible discrete solutions. We establish probabilistic approximation guarantees that characterize the trade-off between computational efficiency, sampling complexity, and curvature of the objective function. Furthermore, we show that under additive transportation costs, the joint allocation-routing problem admits a matroid decoupling, reducing the routing component to shortest-path computations while preserving optimality guarantees. This result provides key theoretical insight into when multi-domain optimization can be simplified without loss of performance. Numerical experiments on synthetic transportation networks demonstrate that COCO-Greedy achieves significant computational gains compared to classical submodular and partition-based approaches, while maintaining competitive solution quality. Overall, this work establishes a new scalable paradigm for multi-domain submodular.

Keywords: Optimization algorithms, Neural networks, Linear systems
Abstract: We study the convergence of model-based policy gradient for the deterministic, scalar, discounted linear-quadratic regulator when the controller is an overparameterized one-hidden-layer ReLU network without biases. Although the optimal LQR controller is linear, neural parameterization creates a redundant, nonconvex weight space and an asymmetric piecewise-linear controller. We show that this structure can still be analyzed exactly through the two effective gains induced on the positive and negative half-lines. Under suitable random initialization, sufficient width, and a small step size, the model-based policy gradient remains stable, decreases the cost geometrically, and drives the effective gains to the unique optimal scalar LQR gain with high probability.

Keywords: Identification, Neural networks, Machine learning
Abstract: Accurate battery state-of-health (SOH) prediction is critical for safe lithium-ion battery operation, yet remains difficult due to highly nonlinear degradation dynamics. We introduce a novel, lightweight physics-informed reservoir computing (PIRC) method that integrates governing physics without requiring gradient-based optimization. By blending data and physics targets into a single closed-form linear solution, PIRC eliminates the need for intensive backpropagation. In the experimental study, PIRC outperforms the standard reservoir computing (RC) method, the hybrid next-generation RC, and the physics-informed neural network method, demonstrating that it is a lightweight alternative that uses only a minimum number of trainable parameters. PIRC offers a robust, deployable solution for real-time battery monitoring on edge devices.

Keywords: Autonomous robots
Abstract: This paper presents a mass-preserving obstacle avoidance method for microscopic control of very-large-scale robotic (VLSR) systems navigating cluttered environments. In the macro-micro framework for VLSR path planning, a macroscopic planner produces a trajectory of Gaussian mixture model (GMM) distributions, and a microscopic controller moves individual agents to track the evolving distribution. We propose a ray-based transport transform that corrects each agent’s position independently by projecting the multivariate problem onto a one-dimensional ray from the assigned component mean. Along this ray, obstacle intersections define a forbidden interval that is excised from the projected Gaussian CDF; the CDF is then re-normalized and the agent’s quantile is preserved via an inverse transform. We show that this directional exclusion is consistent with the BRF-SN framework recently proposed for VLSR systems. The transform is fully decoupled across agents, yielding per-step complexity that is linear in the number of agents.

Keywords: Fault detection, Fault diagnosis
Abstract: This paper presents a novel approach for fault classification in rotating machinery by integrating physics-informed neural networks (PINNs) with machine learning techniques. Our method employs an inverse PINN to estimate speed-dependent dynamic parameters—specifically damping and stiffness coefficients—from cycle-segmented vibration data. By embedding the governing differential equations of rotor dynamics directly into the neural network training, the proposed framework ensures physical consistency and improves the estimated parameters' interpretability. These parameters are subsequently used as discriminative features in a random forest classifier to distinguish between normal operation and misalignment faults. Simulations on a rotary machine dataset demonstrate that the estimated parameters capture meaningful variations across different operating conditions, and the classifier achieves highly robust fault diagnosis, even when subjected to minor variations in misalignment severity.

Keywords: Automotive control, Automotive systems
Abstract: Eco-driving strategies for automated vehicles typically optimize vehicle-level power demand, but often neglect internal powertrain dynamics such as gear shifting, energy conversion, and hybrid energy routing. As a result, identical driving demands can produce different energy outcomes across electric vehicles (EVs), internal combustion engine vehicles (ICEVs), and hybrid electric vehicles (HEVs). This work presents a computationally efficient Koopman-based model predictive control (MPC) framework that explicitly incorporates powertrain-aware dynamics into real-time eco-driving. A unified car-following scenario with an FTP-75-based leader ensures consistent traffic conditions across vehicle types. Koopman-based lifted models enable tractable optimization while capturing key nonlinear powertrain behaviors. Vehicle-specific objectives reflect distinct energy pathways: EVs optimize electricity usage and smoothness, ICEVs minimize fuel consumption and torque variation, and HEVs balance fuel-electric trade-offs with battery state-of-charge regulation. A two-level analysis first isolates powertrain-dependent energy flows under a shared baseline, then quantifies how optimized trajectories modify internal loss mechanisms. Results show that even under identical driving demands, energy consumption is governed by powertrain-specific loss distributions. EVs are dominated by conversion and charge-discharge cycling losses, ICEVs by fuel and braking losses, and HEVs by engine–motor energy routing. The proposed approach achieves energy savings of 4.49% (EV), 6.09% (ICEV), and 9.50% (HEV) with real-time computational performance. Trajectory smoothing consistently reduces unnecessary acceleration and braking, but final savings are determined by how each powertrain processes and distributes energy. These results show that eco-driving performance emerges from the interaction between driving behavior and internal energy mechanisms. This work provides a tractable framework for powertrain-aware eco-driving and insights for energy-efficient automated mobility systems.

Keywords: Agents-based systems
Abstract: We propose a four-agent framework for operating complex swarm robotic systems in both virtual and real environments. The Interface Agent connects user intent with laboratory resources, manages physical and virtual I/O, and coordinates communication among agents. The Algorithm Agent develops task-specific control, planning, and coordination strategies, while also generating low-cost proof-of-concept experiments. The Testing Agent validates algorithm correctness, safety, and robustness under controlled conditions. The Training Agent deploys validated methods into simulation and learning environments, optimizes performance, and returns feedback when errors or instability arise. Together, these four agents form a closed-loop system that supports idea expansion, rapid prototyping, validation, training, and deployment for swarm robotics. In addition to their functional roles, all four agents are connected through a shared Main Wisdom Memory, which serves as a centralized repository of accumulated knowledge, including user intent, safety constraints, experiment history, resource availability, and previously successful or failed strategies. Alongside this shared memory, each agent maintains its own local memory to preserve role-specific knowledge. For example, the Interface Agent stores configuration and resource context, the Algorithm Agent stores design and planning history, the Testing Agent stores validation and failure patterns, and the Training Agent stores optimization trajectories and policy evolution. This dual-memory structure enables both collective learning across the full system and specialized adaptation within each agent.

Keywords: Smart grid, Distributed control, Reinforcement learning
Abstract: This work presents a distributed cooperative secondary control framework for islanded inverter-based microgrids with simultaneous voltage synchronization and state-of-charge (SoC) balancing. Each distributed generator (DG) is modeled to have higher-order nonlinear dynamics, and the voltage synchronization problem is transformed through feedback linearization into a distributed tracking problem. To incorporate battery energy dynamics, the standard DG model is augmented with SoC-related states, enabling coordinated energy management in addition to secondary voltage control. A distributed SoC-balancing layer is then designed, where the SoC control gains are learned using an integral reinforcement learning algorithm. Simulation results show that the proposed method achieves fast terminal-voltage synchronization, stable neighbor-to-neighbor coordination, and improved SoC balancing among all DGs. These results demonstrate the potential of integrating learning-based distributed balancing and secondary control for resilient and energy-aware microgrid operation.

Keywords: Energy systems, Estimation, Observers for nonlinear systems
Abstract: State of charge (SoC) cannot be measured directly in real time; only terminal voltage and current are available from the battery. This work uses measured voltage-current data to first identify the parameters and capacity of a nonlinear 2-RC equivalent circuit model (ECM) through adaptive estimation. A nonlinear observer is then developed for SoC estimation. Results show that the adaptive identification enables accurate battery modeling that drives the nonlinear observer towards achieving faster convergence. Comparisons with a baseline extended Kalman filter (EKF) are carried out with parametric uncertainties to demonstrate effectiveness of the proposed design.

Keywords: Energy systems, Modeling
Abstract: Modern energy systems in vehicles and built infrastructure are governed by high-dimensional dynamics spanning multiple physical domains (e.g., electrical, thermal, mechanical) and timescales. This tutorial paper presents a graph-based modeling approach created to facilitate the modeling, analysis, control, estimation, optimization, and design of these systems. Matured and validated through more than a decade of research spanning multiple academic institutions and companies, the graph-based approach combines transient energy conservation with an explicit mathematical representation of the network by which energy is stored and transferred within a system. Following a mathematical overview of graph-based models, examples of multi-domain component and system models from the recent literature are presented, including single-phase thermal systems, two-phase thermal systems, and electro-mechanical systems. This is followed by a survey of recent applications for decentralized and hierarchical model predictive control, design optimization, and control co-design. Lastly, the paper describes an open-source toolbox created to facilitate the generation and analysis of graph-based models.

Keywords: Spacecraft control, Lyapunov methods, Kalman filtering
Abstract: Recent advancements in space exploration have increased the number of missions in the cislunar region, highlighting the growing importance of reliable rendezvous and proximity operations (RPO). This paper presents a unified framework that integrates dynamics, estimation, and control for safe and precise cislunar RPO. The translational and rotational motions of the spacecraft are modeled using the circular restricted full three-body problem (CRF3BP), which captures coupled attitude–orbit dynamics. To guarantee both stability and safety, a control Lyapunov-barrier function (CLBF) based controller is developed. To estimate the target’s relative state, LiDAR point cloud data are processed using a generalized iterative closest point (G-ICP) algorithm. The proposed end-to-end algorithm is validated through numerical simulations, demonstrating its effectiveness in ensuring accurate state estimation and safe trajectory tracking for cislunar RPO missions.

Keywords: Kalman filtering, Algebraic/geometric methods, Estimation
Abstract: The unscented Kalman filter (UKF), while effective in many applications, relies on a symmetric Gaussian assumption that limits its ability to capture complex, non-Gaussian uncertainty. Although the unscented transform itself does not require Gaussian noise, the standard UKF formulation does not retain this flexibility. A Gaussian mixture extension of the square-root UKF (SR-UKF), developed directly on the special Euclidean group SE(3) and its tangent bundle TSE(3), is introduced to preserve the geometric structure of rigid body motion while capturing multimodal posteriors under limited attitude observability. Multiple SR-UKFs (mixture components) are propagated in parallel, where the weighted sum of these filters enables an effective approximation of the nonlinear state evolution. Performance is demonstrated through numerical simulation of spacecraft pose and velocity estimation in a near-rectilinear halo orbit in the cislunar environment with limited observability.

Keywords: Algebraic/geometric methods, Computational methods, Aerospace
Abstract: In this work, we propose a novel approach, based on the path integral formulation of Koopman spectrum, to discover the phase-space geometry of the planar Cislunar Restricted 3 Body Problem (CR3BP). In contrast to existing techniques, which use trajectory-based (usually) local analysis, we leverage the Koopman operator framework, which generates a global linear emph{representation} of the system, to reconstruct the global phase space geometry of the CR3BP. In particular, we compute the principal eigenfunctions of the Koopman operator via the path integral approach and show how the zero level curves of these eigenfunctions encode the phase space characteristics of the planar CR3BP.

Keywords: Optimal control, Spacecraft control
Abstract: This work presents a novel algorithm for impulsive optimal control of linear time-varying systems with the inclusion of input magnitude constraints. Impulsive optimal control problems, where the optimal input solution is a sum of delta functions, are typically formulated as an optimization over a normed function space subject to integral equality constraints and can be efficiently solved for linear time-varying systems in their dual formulation. In this dual setting, the problem takes the form of a semi-infinite program which is readily solvable in online scenarios for constructing maneuver plans. This work augments the approach with the inclusion of magnitude constraints on the input over time windows of interest, which is shown to preserve the impulsive nature of the optimal solution and enable efficient solution procedures via semi-infinite programming. The resulting algorithm is demonstrated on the highly relevant problem of relative motion control of spacecraft in Low Earth Orbit (LEO).

Keywords: Direct adaptive control, Spacecraft control, Uncertain systems
Abstract: This paper presents an adaptive backstepping controller formulated on the special orthogonal group SO(3), designed to address uncertainties in the spacecraft’s inertia tensor without relying on gain scheduling. Taking advantage of the intrinsic Lie group-based attitude representation, the pro- posed controller provides adaptability to poorly modeled phe- nomena such as refueling, structural deployment, or damages. The effectiveness of the proposed controller is demonstrated both theoretically, through Lyapunov-based stability analysis, and numerically, via a simulation involving time-varying and unknown inertia, showing successful tracking of the desired reference attitude despite these difficulties. A comparison with a geometric PD tracking controller that includes feedforward and dynamic inversion demonstrates that the proposed adaptive approach is more accurate when the effort is comparable.

Keywords: Game theory, Spacecraft control, Optimization
Abstract: We develop a framework for computing best response actions of two competitive agents in proximity to each other in low-earth orbit; one agent remains fixed at the origin while the other approaches according to Clohessy-Wiltshire-Hill dynamics. We characterize best-response solutions for the ap- proaching agent dependent on the orientation of the fixed agent by solving a linear quadratic optimization problem for a set of waypoints and open-loop controls for the approaching agent. We propose efficient homotopy-based methods for computing all optimal solutions for convex combinations of possible costs for the approaching agent. The fixed agent can then use these best response maps to choose an orientation that reveals different intentions of the approaching agent in a manner that is robust to variation in the approaching agent’s cost. We demonstrate in simulation that these techniques can effectively distinguish between imaging objectives for the approaching agent.

Keywords: Game theory, Control of networks, Optimization
Abstract: Resource allocation is a fundamental challenge in control theory, with applications spanning security, surveillance, network control, and market strategy. These problems are often associated with decision-making in adversarial environments, where players compete for control over valued contests. When adversaries are strategic themselves, how should one allocate limited resources to maximize their individual objectives? This tutorial paper serves as a companion summary of the session presented at the 2026 American Control Conference. We examine the emergence of "Colonel Blotto games", a class of game-theoretic models that aptly represents such scenarios. This paper briefly showcases a collection of recent works that have made fundamental contributions in the analysis of Blotto games and their variants. They have enabled versatile applications to networked control, cybersecurity, economic competitions, political contests, and defense. We direct the interested reader to a full in-depth expository article on this subject.

Keywords: Networked control systems, Control over communications, Decentralized control
Abstract: This letter explores intelligent scheduling of sensor-to-controller communication in networked control systems, particularly when data transmission incurs a cost. While the optimal controller in a standard linear-quadratic Gaussian (LQG) setup can be computed analytically, determining the optimal times to transmit sensor data remains computationally and analytically challenging. We show that, through reformulation and the introduction of auxiliary binary variables, the scheduling problem can be cast as a computationally efficient mixed-integer linear program (MILP). This formulation not only simplifies the analysis but also reveals structural insights and provides clear decision criterion at each step. Embedding the approach within a model predictive control (MPC) framework enables dynamic adaptation, and we prove that the resulting scheduler performs at least as well as any deterministic strategy (e.g., periodic strategy). Simulation results further demonstrate that our method consistently outperforms traditional periodic scheduling.

Keywords: Networked control systems, Distributed control, Power electronics
Abstract: This paper presents a dissipativity-based distributed control and topology co-design approach for DC microgrids (DC MGs) with ZIP loads. We address challenges from ZIP loads, particularly their nonlinear and destabilizing constant power loads (CPLs), while ensuring robust voltage regulation and current sharing. We model the DC MG as a networked system with distributed generators (DGs), ZIP loads, and transmission lines, and propose local and distributed controllers for each DG. By formulating closed-loop error dynamics as a networked system and using sector-bounded CPL characterization, we develop a computationally efficient one-shot co-design framework solved as a linear matrix inequality (LMI), which embeds necessary conditions locally while ensuring global feasibility, avoiding iterative synthesis procedures. The simulation results demonstrate superior performance compared to the traditional approaches.

Keywords: Networked control systems, Network analysis and control, Linear systems
Abstract: We develop a data-driven framework for assessing the resilience of linear time-invariant systems against malicious false-data-injection sensor attacks. Leveraging sparse observability, we propose data-driven resilience metrics and derive necessary and sufficient conditions for two data-availability scenarios. For attack-free data, we show that when a rank condition holds, the resilience level can be computed exactly from the data alone, without prior knowledge of the system parameters. We then extend the analysis to the case where only poisoned data are available and show that the resulting assessment is necessarily conservative. For both scenarios, we provide algorithms for computing the proposed metrics and show that they can be computed in polynomial time under an additional spectral condition. A numerical example illustrates the efficacy and limitations of the proposed framework.

Keywords: Networked control systems, Optimization, Optimization algorithms
Abstract: This paper addresses a distributed nonconvex optimization problem over multi-agent networks, where each agent exchanges its local information solely with its neighbors. Given that most existing distributed nonconvex optimization algorithms are susceptible to information leakage during inter-agent communications, we propose a Robust Proximal Primal-dual algorithm, referred to as RPP, to enhance the security of information transmission. In contrast to many existing approaches that directly transmit local variables throughout the network, we introduce carefully designed random noises to obfuscate sensitive local information. This not only preserves privacy but also demonstrates the noise robustness of our proposed algorithm. We establish a sublinear rate at which RPP converges to a stationary solution. Moreover, by incorporating Chebyshev acceleration, an accelerated variant of RPP is developed and achieves the optimal communication complexity bound for the algorithms that allow for exchanging local decisions at each iteration. The superior convergence performance of RPP is validated through a few numerical experiments, which also indicate that, within an appropriate range, the introduced perturbations do not impede the convergence speed of RPP.

Keywords: Networked control systems, Optimization algorithms
Abstract: Cellular sheaves and sheaf Laplacians provide a far-reaching generalization of graphs and graph Laplacians, resulting in a wide array of applications ranging from machine learning to multi-agent control. In the context of multi-agent systems, so called coordination sheaves provide a unifying formalism that models heterogeneous agents and coordination goals over undirected communication topologies, and applying sheaf diffusion drives agents to achieve their coordination goals. Existing literature on sheaf diffusion assumes that agents can communicate and compute updates synchronously, which is an unrealistic assumption in many scenarios where communication delays or heterogeneous agents with different compute capabilities cause disagreement among agents. To address these challenges, we introduce asynchronous nonlinear sheaf diffusion. Specifically, we show that under mild assumptions on the coordination sheaf and bounded delays in communication and computation, nonlinear sheaf diffusion converges to a minimizer of the Dirichlet energy of the coordination sheaf at a linear rate proportional to the delay bound. We further show that this linear convergence is attained from arbitrary initial conditions and the analysis depends on the spectrum of the sheaf Laplacian in a manner that generalizes the standard graph Laplacian case. We provide several numerical simulations to validate our theoretical results.

Keywords: Nonlinear systems identification, Identification, Networked control systems
Abstract: In this work, we study the identifiability of network structures (i.e., topologies) for networked nonlinear systems when partial measurements of the nodal dynamics are taken. We explore scenarios where different candidate structures can yield similar measurements, thus limiting identifiability. To do so, we apply the contraction theory framework to facilitate comparisons between different networks. We show that semicontraction in the observable space is a sufficient condition for two systems to become indistinguishable from one another based on partial measurements. We apply this framework to study networks of Kuramoto oscillators, and discuss scenarios in which different network structures (both connected and disconnected) become indistinguishable.

Keywords: Automotive control, Adaptive control, Stochastic optimal control
Abstract: Vehicle motion planning critically depends on traction, yet real-time estimation of tire-terrain interaction remains a fundamental challenge in autonomous driving. Traditional approaches rely on passive estimation, where friction is inferred indirectly as a consequence of trajectory tracking. Such methods often fail when system dynamics do not provide sufficient excitation, leaving friction unobservable thereby degrading performance. To overcome this limitation, we propose ADAPT - Adaptive Dual control for Active Planning under Traction uncertainty, a novel dual-stochastic Model Predictive Path Integral (MPPI) framework that actively reduces uncertainty while optimizing control objectives. This study incorporates both implicit and explicit dual control formulations: the implicit variant triggers exploration only when it improves performance, while the explicit variant deliberately excites the system under high uncertainty. Through a simulated left-turn scenario with uncertain friction, we demonstrate that both dual-MPPI formulations reduce uncertainty more effectively and outperform a certainty-equivalent MPPI baseline, with the implicit approach achieving a more efficient exploration-exploitation tradeoff. These results highlight the potential of dual control to enhance safety, adaptability, and performance of autonomous systems operating in uncertain traction environments.

Keywords: Automotive control, Predictive control for nonlinear systems, Fluid power control
Abstract: This paper presents a model predictive energy management strategy for a parallel electric-hydraulic hybrid vehicle, aimed to improve energy efficiency via hydraulic regenerative braking. A physics-based mathematical model is first derived and validated against a high-fidelity multiphysics simulation model. Building on this, a model-predictive energy management strategy with a finite look-ahead horizon is formulated to optimize high-level torque distribution between the electric and hydraulic components. The proposed strategy is numerically validated and compared with a global optimization solution obtained via dynamic programming. The comparative analysis highlights both the potential and the limitations of the proposed method. The model predictive energy management strategy is computationally efficient and achieves near-optimal energy efficiency. However, it generally involves more intensive operation of the electric motor and mechanical brakes, which may lead to elevated electrical stress on the battery and accelerated wear of the braking system.

Keywords: Automotive control, Predictive control for nonlinear systems, Optimal control
Abstract: This paper introduces a novel Raceline-Informed Envelope-Based Model Predictive Control (RE-MPC) framework for autonomous racing that pushes the vehicle to its dynamic limits through real-time planning while accounting for an online-generated geometric raceline. Prior work has shown that Envelope-Based Model Predictive Control (E-MPC) can generate dynamically feasible and collision-free racing trajectories without relying on a predefined reference path. However, its local planning nature may result in suboptimal behavior in complex scenarios. To mitigate this shortcoming of E-MPC, the new RE-MPC framework incorporates a reference raceline into the optimization process. The approach consists of two steps. First, similar to the state of the art, the drivable region for each track is segmented into a sequence of blocks that collectively form an envelope, ensuring that the vehicle remains within track boundaries. The second step marks the original contribution of this paper by generating a geometric raceline online and applying additional costs on selective prediction horizon points along the raceline rather than strictly adhering to it. The RE-MPC framework is evaluated and validated in simulation across various race tracks for autonomous racing. The results demonstrate that RE-MPC improves racing performance over the state of the art on all the tracks tested while maintaining real-time computational performance. These results indicate the potential of the RE-MPC approach to provide performant, robust, and real-time feasible optimization solutions for autonomous racing in diverse scenarios.

Keywords: Automotive control
Abstract: Model-free control (MFC) offers a simple and effective approach to automated vehicle path-tracking without requiring an explicit plant model for control law design. However, gain tuning in MFC is typically carried out through trial-and-error, which can be time-consuming and may lead to suboptimal performance. To address this limitation, extremumseeking- based adaptive MFC has shown promise by enabling real-time adaptation of control gains, without relying on a predefined vehicle model. Nonetheless, existing ESC approaches often suffer from slow convergence. This paper integrates MFC, employing longitudinal and lateral ultra-local models of a rear-wheel-drive vehicle, with a novel prescribed-time (PT) Newton-based extremum-seeking control (ESC) strategy that ensures rapid convergence of control gains within the prescribed time. Unlike conventional gradient-based ESC methods, the PT Newton-based ESC leverages artificial delays and time-periodic gains, not only to guarantee convergence within the specified time, but also to compensate for feedback delays. Simulation results demonstrate that the proposed approach significantly improves gain adaptation speed and tracking accuracy. This work advances adaptive model-free vehicle control by offering a high-performance, delay-resilient alternative to existing ESCMFC frameworks.

Keywords: Automotive control, Reinforcement learning
Abstract: Efficient energy and emission management in heavy-duty hybrid electric vehicles requires controllers that balance near-optimal performance with real-world practicality. Interpretable Reinforcement Learning (IRL) addresses this need by jointly optimizing fuel economy and tailpipe emissions while providing decision tree (DT) and lookup table (LUT) structures that enable traceable and refinable control logic. Stochastic Dynamic Programming (SDP) is used as a benchmark, revealing that although it increases nitrogen oxide (NO) emissions by 3.10%, it improves fuel economy by 10.73%, achieving an overall advantage of 3.45% under a combined reward function. Policy-level analysis shows SDP’s superior robustness under extreme power demands, where IRL may risk state-of-charge violations, while their behavior remains highly consistent under typical conditions. Moreover, insights from SDP can guide manual refinement of IRL policies without retraining, reducing performance gaps: relative to the refined IRL, SDP shows only 1.94% higher NO emissions and a 6.41% gain in fuel economy, for an overall advantage of 2.02%. These results demonstrate that IRL can closely approximate optimal performance while remaining interpretable, editable, and scalable, and that IRL enhances its robustness by leveraging SDP as both a benchmark and a policy-level oracle.

Keywords: Automotive control, Reinforcement learning, Neural networks
Abstract: This study presents an Actor-Critic Compensated Model Predictive Controller designed for high-speed, off-road autonomous driving on deformable terrains. Addressing the difficulty of modeling unknown tire-terrain interaction and ensuring real-time control feasibility and performance, this framework integrates deep reinforcement learning with a model predictive controller to manage unmodeled nonlinear dynamics. We evaluate the controller framework over constant and varying velocity profiles using high-fidelity simulator Project Chrono. Our findings demonstrate that our controller statistically outperforms standalone model-based and learning-based controllers over three unknown terrains that represent sandy deformable track, sandy and rocky track and cohesive clay like deformable soil track. Despite varied and previously unseen terrain characteristics, this framework generalized well enough to track longitudinal reference speeds with the least error. Furthermore, this framework required significantly less training data compared to purely learning based controller, converging in fewer steps while delivering better performance. Even when under-trained, this controller outperformed the standalone controllers, highlighting its potential for safer and more efficient real-world deployment.

Keywords: Traffic control, Reinforcement learning, Decentralized control
Abstract: Traffic congestion in urban areas causes delays, increased emissions, and economic losses. Existing traffic signal control systems struggle to adapt to dynamic conditions, resulting in inefficiencies and frequent gridlock. This paper introduces Multi-Dancer, a novel framework within Multi-Agent Reinforcement Learning that pioneers strategies specifically designed to mitigate congestion. Multi-Dancer features adaptive team formation, enabling traffic signal controllers to dynamically adjust coordination based on relevant information from neighboring intersections, and surrogate experience replay, which enhances learning efficiency by managing delayed reward feedback. Our evaluations demonstrate that Multi-Dancer significantly reduces congestion and recovery times by 65% compared to uncoordinated RL methods, surpassing state-of-the-art solutions by cutting average waiting time by over 91% and travel time by over 82% in high-density traffic.

Keywords: Multivehicle systems, Decentralized control, Autonomous robots
Abstract: Decentralized trajectory planning for multiple heterogeneous vehicles in cluttered and uncertain environments remains a significant challenge due to the dual requirements of efficiency and safety. In this paper, we propose a decentralized and asynchronous multi-vehicle trajectory planning framework that leverages risk assessment and chance constraints. First, a risk-aware local target selection strategy and a risk-aware objective are designed to incorporate the risks posed by other vehicles and obstacles into optimization, effectively preventing deadlocks and improving coordination performance. In addition, inter-vehicle and vehicle-obstacle collision avoidance is formulated as chance-constrained problems and then reformulated into tractable deterministic conditions, thus ensuring safety under Gaussian noises. Numerical simulations in static and dynamic environments with heterogeneous vehicles demonstrate that the proposed method achieves higher success rates, larger safety margins, and fewer collisions compared to both the baseline and ablated versions, validating the robustness and effectiveness of the proposed framework.

Keywords: Multivehicle systems, Traffic control, Autonomous systems
Abstract: We consider interchange lane-swap scenario, a limited stretch of highway with two parallel lanes where most vehicles want to change lanes. We show that a particular decentralized Control Barrier Function based algorithm executes lane swaps efficiently, with minimal speed change, and within the specified (short) road segment at high traffic density (3,500 veh./hour per lane). Our main point is that controller tuning – the speed of inter-agent instability – plays a major role in the performance of the vehicle group. This is illustrated by comparing two different tunings of the controller and a third one where the lane swap is enforced by “virtual guard rails.” Like fighter-jet dynamic instability improving maneuverability, in this work the inter-agent instability improves agility of a group of vehicles. We emphasize that the controllers are decentralized: agents do not know if others want to change lanes or not.

Keywords: Multivehicle systems, Automotive systems, Autonomous systems
Abstract: This paper presents a decentralized Control Barrier Function (CBF) based approach for highway merging of Connected and Automated Vehicles (CAVs). In this control algorithm, each "host" vehicle negotiates with other agents in a control zone of the highway network, and enacts its own action, to perform safe and energy-efficient merge maneuvers. It uses predictor-corrector loops within the robust CBF setting for negotiation and to reconcile disagreements that may arise. There is no explicit order of vehicles and no priority. A notable feature is absence of gridlocks due to instability of the inter-agent system. Results from Monte Carlo simulations show significant improvement in the system-wide energy efficiency and traffic flow compared to a first-in-first-out approach, as well as enhanced robustness of the proposed decentralized controller compared to its centralized counterpart.

Keywords: Automotive systems, Automotive control, Traffic control
Abstract: Improving the energy efficiency of automated vehicles is essential for sustainable intelligent transportation systems. Energy-optimal driving, which adjusts speed to minimize energy consumption while respecting safety and traffic rules, has benefited from connected and automated vehicle (CAV) technologies that provide signal phase and timing (SPaT) data and upstream traffic information. However, existing model predictive control (MPC) methods are limited by nonlinear powertrain dynamics, while learning-based approaches, though flexible, often struggle with data efficiency, sim-to-real transfer, and formal safety guarantees. This paper presents a unified control framework that combines queue-informed prediction with a Koopman operator based optimizer to enable efficient energy-optimal control in urban corridors. A virtual leader abstraction is introduced, where the ego vehicle's admissible distance is bounded by the nearest limiting factor, including a physical leader, a standing queue, or a signal stop line. A prediction module uses SPaT and queue evolution to estimate these bounds, which are supplied to an optimal control layer formulated as either the prposed Koopman-lifted quadratic programming or a more complex and accurate nonlinear MPC baseline. Closed-loop simulations with a high-fidelity Simulink vehicle model demonstrate that the Koopman-based approach achieves similar energy savings to nonlinear MPC while significantly reducing computation time. The proposed framework provides a scalable foundation for energy-optimal control in realistic traffic environments.

Keywords: Reinforcement learning, Cooperative control, Automotive control
Abstract: Highway on-ramp merging is a safety-critical and efficiency-sensitive maneuver, especially in mixed traffic where connected automated vehicles (CAVs) and human-driven vehicles coexist. Recent advances in multi-agent reinforcement learning (MARL) have demonstrated promising performance in cooperative merging tasks. However, existing reward functions typically oversimplify safety concerns and neglect fuel consumption. In this paper, we propose an enhanced MARL framework for mixed-traffic on-ramp merging that integrates a risk-sensitive safety metric and energy-aware driving objectives. Specifically, an artificial potential field (APF) model is introduced into the reward design to comprehensively penalize collision risks over distance-based metrics, while the VT-Micro microscopic fuel consumption model is incorporated to account for energy efficiency. Simulation experiments, benchmarked against the baseline reward design from recent state-of-the-art MARL studies, demonstrate that the proposed reward shaping yields significantly lower collision rates, higher traffic throughput, and reduced fuel consumption.

Keywords: Fault detection, Fault diagnosis
Abstract: Early detection of internal short circuits (ISCs) is critical for ensuring lithium-ion battery safety across all chemistries. In this work, we investigate the added value of cell-level temperature sensing in conjunction with voltage sensing. We derive the analytical relationship for the time required for a detectable voltage drop and temperature rise during ISC faults for LFP and NMC chemistries, where ``detectable" is defined as the level of fault signatures reaching right above the sensor noise. We then compare our analytical expressions for the rate of voltage drop (dV/dt) and rate of temperature increase (dT/dt) signatures with numerical simulations. We demonstrate that for both chemistries, thermal signatures build up faster than electrical signatures. Under hard shorts, temperature can rapidly confirm the occurrence of an ISC, occurring in under 2 minutes for LFP and under 30 seconds for NMC. For soft shorts, temperature surpasses the noise threshold much earlier than voltage, by 10 to 100 times for LFP and 2 to 4 times for NMC. These results demonstrate that while voltage sensing remains indispensable, cell-level temperature sensing provides earlier detection, independent confirmation, and stronger diagnostic certainty.

Keywords: Energy systems, Multivehicle systems, Optimization
Abstract: With the number of electric vehicles (EVs) in the San Francisco Bay Area projected to grow from 0.5 million to around 4 million in the next decade, uncoordinated charging poses a significant threat to power infrastructure. This load growth, under current charging behavior patterns, is estimated to require billions of dollars in infrastructure upgrades by 2035, costs that are unevenly distributed across the region's 1000+ distribution feeders. To complement hardware upgrades, we explore the full potential of leveraging EV demand flexibility. Our central question is: what is the maximum potential of a holistically coordinated charging system to mitigate feeder overloading risks?

We propose a ``mobility-aware'' charging coordination framework that guarantees EVs maintain feasible energy levels throughout their week-long trajectories, a significant departure from conventional models that set energy demand constraints for each individual charging session. To tackle the computationally intractable scale of this problem, we developed a custom Alternative Direction Method of Multipliers (ADMM) approach for efficient distributed optimization on parallel clusters. Using realistic EV trajectory data and feeder-level hosting capacity, we assess the maximum potential for mitigating overloading events across the Bay Area. By comparing with rule-based or individually optimized charging strategies, the fully coordinated scheme can reduce total feeder upgrade requirements by approximately 6 billion USD.

Keywords: Energy systems, Modeling
Abstract: The solid electrolyte interphase (SEI) – a critical passivation layer that governs the longevity, safety, and efficiency of lithium-ion batteries – is created during the last step in cell manufacturing called cell formation. Conventional cell formation protocols are largely empirical, resulting in long processing times and limited control over the SEI growth rate that influences SEI quality and lifetime performance. This paper develops a control-oriented, semi-empirical model to estimate SEI thickness growth from terminal voltage and cell expansion measurements acquired in-operando during manufacturing using low-cost micrometer-precision integrated-sensing fixture. Model parameters are calibrated against cell formation data, and an unscented Kalman filter is employed to estimate the SEI film growth. The results lay the foundation for future closed-loop control of SEI growth, enabling high-quality and more efficient formation processes.

Keywords: Fault detection, Fault diagnosis, Energy systems
Abstract: Early diagnosis of battery thermal anomalies is crucial to ensure safe and reliable battery operation by preventing catastrophic thermal failures. Battery diagnostics primarily rely on battery surface temperature measurements and/or estimation of core temperatures. However, aging-induced changes in the battery model and limited training data remain major challenges for model-based and machine-learning based battery state estimation and diagnostics. To address these issues, we propose a Kolomogorov-Arnold network (KAN) in conjunction with a Koopman-based detection algorithm that leverages the unique advantages of both methods. Firstly, the lightweight KAN provides a model-free estimation of the core temperature to ensure rapid detection of battery thermal anomalies. Secondly, the Koopman operator is learned in real time using the estimated core temperature from KAN and the measured surface temperature of the battery to provide a prediction for diagnostic residual generation. This online learning approach overcomes the challenges of model changes, while the KAN–Koopman integrated structure reduces the dependence on large datasets. Furthermore, we derive analytical conditions that provide diagnostic guarantees on our KAN-Koopman detection scheme. Our simulation results illustrate a significant reduction in detection time with the proposed algorithm compared to the baseline Koopman-only algorithm.

Keywords: Modeling, Energy systems, Fault diagnosis
Abstract: Electrochemical faults arising from internal degradation-related side reactions in lithium-ion batteries play a crucial role in battery safety and performance. Accurate modeling and thorough understanding of these processes are essential. The side reactions considered in this study are two of the most critical degradation-related electrochemical mechanisms: Solid Electrolyte Interphase (SEI) film growth and lithium plating. The Doyle–Fuller–Newman model with Side Reactions (DFN+SR) and the Single Particle Model with electrolyte dynamics integrated with Side Reactions (SPMe+SR) are first described. Subsequently, reformulated models providing a parameterized representation of side-reaction dynamics suitable for diagnostic analysis are proposed. These models are developed with strong physical justification while offering substantially lower computational cost and complexity. The study is structured in two parts, each standing independently, and Part 2 complements Part 1 by developing fault diagnosis schemes. The proposed reformulated side-reaction models are simulated on a LiFePO4 cell to capture the characteristic degradation behavior.

Keywords: Fault diagnosis, Energy systems, Fault detection
Abstract: This paper presents a novel fault diagnosis framework for lithium-ion batteries in both the electrolyte and solid phases. The proposed scheme enables the detection, estimation, and isolation of electrochemical faults, including SEI film growth and lithium plating. This study consists of two parts, each of which can stand independently. Part 1 introduces the Single Particle Model with electrolyte dynamics integrated with side reactions (SPMe+SR), which serves as the battery model for this study, and proposes reformulated electrochemical fault models that facilitate the development of a fault diagnosis scheme. In this study, the proposed fault diagnosis scheme employs PDE backstepping cascaded observers, where the first estimates the distributed lithium concentration, and the second, an adaptive observer, estimates the fault magnitude via an update law. The effectiveness of the proposed approach is validated through simulations on a LiFePO 4 cell under an Urban Dynamometer Driving Schedule (UDDS) current profile.

Keywords: Delay systems, Lyapunov methods, Neural networks
Abstract: Abstract--- Neurological conditions (NCs) often lead to motor impairments that limit functional independence and quality of life. Repetitive, task-specific movements are critical for promoting neuroplasticity and functional recovery. However, access to in-person rehabilitation is frequently limited by logistical, geographic, and socioeconomic barriers, which has motivated the development of home-based telerehabilitation systems. In a centralized leader–follower telerobotic setup, state information is transmitted from a user's robotic device to a remote (clinical) facility, where control commands are generated and sent back for execution. Communication-induced input delays in the control generation process can degrade tracking performance or destabilize the system, particularly in dynamic rehabilitation tasks that require precise control. For the first time, this paper presents a neural network (NN)- and Lyapunov-based control framework for uncertain nonlinear telerobotic rehabilitation systems with unknown time-varying input delays. Unlike prior telerobotic results, this work uses a centralized framework and employs a NN to adaptively update the controller by approximating the unknown system dynamics online, despite the uncertain delays. Moreover, a Lyapunov-like analysis is employed to establish online training laws for the NN and to rigorously prove semi-global exponential trajectory tracking with convergence to an ultimate bound.

Keywords: Delay systems, Neural networks, Agents-based systems
Abstract: We propose a novel approach to address the reference tracking problem for discrete-time multi-agent systems with delayed inputs using deep neural networks. The tracking problem is considered an optimization problem with a proposed loss function consisting of tracking errors, consensus errors, and control cost. We employ non-fully connected deep neural networks (NFC-DNNs) to obtain optimized control policies for all agents. The tracking error and delayed inputs of agents are utilized as input to NFC-DNNs for producing optimized control policies. For the update of NFC-DNNs' weights in the optimization of control policies, we develop a tailored gradient computation algorithm for the minimization of the loss function. We conduct numerical experiments with comparison to the baseline of predictor feedback. Simulation results show the superiority of our method over predictor feedback in terms of settling time, peak value and tracking error convergence.

Keywords: Delay systems, Robust control
Abstract: It is difficult to analyze the stability of systems with time-varying delays. One approach is to construct a time-transformation that converts the system into a form with a constant delay but with a time-varying scalar appearing in the system matrices. The stability of this transformed system can then be analyzed using methods to bound the effect of the time-varying scalar. One issue is that this transformation is non-unique and requires the solution of an Abel equation. A specific time-transformation typically must be computed numerically. We address this issue by computing an explicit, although approximate, time-transformation for systems where the delay has a constant plus small periodic term. We use a perturbative expansion to construct our explicit solutions. We provide a simple numerical example to illustrate the approach. We also demonstrate the use of this time-transformation to analyze stability of the system with this class of periodic delays.

Keywords: Delay systems, Time-varying systems, Robust control
Abstract: Delay Differential Equations (DDEs) with an uncertain time-varying state delay are often modeled as a feedback interconnection of a nominal time-invariant system with a time-varying uncertainty operator. However, typical choices of uncertainty operators do not admit tight characterizations of input-output properties. In this paper, we propose a new representation of a DDE with time-varying delay system as a feedback interconnection of a nominal fixed delay DDE with a time-varying multiplicative parametric uncertainty operator, where the parameter is bounded in L_infty norms -- yielding an operator norm bound on the uncertainty. We then provide a modified uncertainty representation which uses the L_2 norm of the parameter to bound the operator norm of the uncertainty. Finally, we show that by representing the nominal system using Partial Integral Equations (PIEs), these characterizations of the uncertainty operator can be used for stability analysis or stabilizing feedback controller synthesis. The effectiveness of the approach is illustrated with simple numerical examples.

Keywords: Delay systems, Variable-structure/sliding-mode control, Stability of nonlinear systems
Abstract: We study a general class of nonlinear systems containing unknown nonlinearities, unknown bounded time-varying state delays, and unknown perturbations of the nonlinearities, using sliding mode event-triggered controls. We provide new sufficient conditions on the delays and perturbations that ensure finite-time convergence to an arbitrarily small forward-invariant neighborhood of a sliding surface, and global asymptotic stabilization to a prescribed neighborhood of an equilibrium. We illustrate the effectiveness of our new analysis using an inverted pendulum dynamics

Keywords: Observers for Linear systems, Delay systems, Linear systems
Abstract: This paper investigates an unknown input observer for time delay systems for the Strong Observable and Strong Detectable cases. Conditions are derived to transform the system into an observer form and to replace the unknown inputs. This replacement allows us to use the measurements obtained at the output of the system to compensate for the missing information. These conditions enable the design of a Luenberger-like observer for systems Strongly Observable and Strongly Detectable, thereby relaxing the condition for this type of systems.

Keywords: Mechanical systems/robotics, Lyapunov methods, Algebraic/geometric methods
Abstract: This paper addresses the problem of 3D coverage control coupled with object alignment for a given mission domain. We utilize an Unmanned Ground Vehicle (UGV) equipped with a camera capable of rotating in the yaw, pitch, and roll directions. In contrast to existing works, we develop system models that include the coupled kinematics of the UGV and camera defined on the Lie group SE(3) and the position kinematics of a point in the sensor domain defined in the camera frame. Then, an appropriate sensor model is formulated by defining a Lie algebra function to consider distance-based coverage and object alignment in a single function. Providing such a unified framework allows us to design control laws for the UGV and camera to achieve 3D dynamic coverage control and object alignment simultaneously while also respecting the coupled kinematics of the underlying SE(3) motion of the camera and UGV. We define a Lyapunov error as a convex function of the total coverage accumulated by each point in the mission domain. The control laws derived are asymptotically stable, proposed under the framework of SE(3), analytic and guarantee coverage of the entire domain and ensure alignment of the camera with objects of interest. Simulation results are shown for convergence of the Lyapunov error function and snapshots of coverage achieved at different time steps.

Keywords: Autonomous robots, Reinforcement learning, Network analysis and control
Abstract: Efficient information diffusion is critical for coordination in large-scale robot swarms operating under dynamic communication topologies and limited energy resources. This paper studies diffusion control by modeling information propagation as a node-level susceptible–infected–recovered (SIR) process and formulating a finite-horizon optimal control problem that jointly regulates transmission intensity and robot mobility policies. We first derive a spectral diffusion condition with a time window that characterizes when information spreading is sustained in time-varying networks, and establish input-to-state stability (ISS) bounds that reveal a fundamental trade-off between diffusion speed and robustness. These theoretical insights motivate a theory-guided multi-robot reinforcement learning framework termed SIR-informed multi-agent proximal policy optimization (SIR-MAPPO). The framework adopts centralized training with decentralized execution and embeds feasibility projections, Lyapunov-informed critics, and hybrid model-based/model-free gradients to enforce stability and bandwidth constraints during learning. Convergence and stability of the learning dynamics are established analytically. Simulations of large-scale robot swarms demonstrate that the proposed approach achieves faster information diffusion and lower communication energy than representative baselines.

Keywords: Autonomous robots, Optimization, Cooperative control
Abstract: In the context of social navigation, a robot must follow social norms while navigating towards its goal. To capture the environment around the robot, existing algorithms model pedestrians as independent agents. However, in densely crowded situations, this approach can be computationally expensive and adversely impact the robot's performance. In this work, we explore ways to achieve better performance and well-behaved scaling properties in such situations. We discuss how pedestrians navigate around groups, focusing on textit{group cohesion} and textit{group intrusion avoidance}. We propose a suitable mathematical model for approximating an enclosure of the group, as well as motion of the group and its enclosure. We evaluate our approach by augmenting it on various existing algorithms and demonstrate that our group approximation method not only improves computational efficiency via dimensionality reduction, but also maintains social appropriateness.

Keywords: Autonomous robots, Cooperative control, Multivehicle systems
Abstract: This paper proposes Multi-UAV Flow Matching with Residual Policy Learning (MUFM-RPL), a distributed control framework for ergodic coverage in collaborative building reconstruction. Classical Fourier-based ergodic methods scale poorly and often generate redundant trajectories in multi-robot settings. To address this issue, we recast ergodic coverage as a distribution alignment problem through flow matching. The resulting trajectory synthesis admits a block-structured linear quadratic regulator (LQR) formulation. Exploiting this structure, we develop a distributed Riccati solver based on the Alternating Direction Method of Multipliers (ADMM) for scalable coordination under communication constraints. To improve robustness to model uncertainty and disturbances, the controller is embedded in a receding-horizon Model Predictive Control (MPC) framework and augmented with RPL. We further establish Lyapunov-based convergence guarantees for the closed-loop system under bounded residual policies. Extensive simulations show that MUFM improves reconstruction fidelity, coverage efficiency, and robustness over existing ergodic and reinforcement-learning baselines.

Keywords: Autonomous robots, Constrained control, Robust control
Abstract: This paper studies the problem of safety-critical motion control in the presence of multiple dynamic obstacles. Based on the existing results in Environmentally Robust Control Barrier Function (ER-CBF), we present some extensions and generalizations to the framework. Specifically, the proposed ERCBF-QP framework in this paper applies to multiple dynamic obstacles and a multi-dimensional control space. With our controller, the mobile robot can successfully avoid multiple dynamic obstacles despite measurement errors in environmental states. The simulation result helps verify the effectiveness of our method for the case of avoiding multiple pedestrians.

Keywords: Autonomous robots, Adaptive control, Sensor networks
Abstract: In this work, we address the problem of multi-robot adaptive coverage, where teams of robots perform dynamic sampling by continuously adjusting their positions to collect data in an environment. This task can be challenging, particularly when robots must be efficiently allocated to new sampling locations over time. Ergodic search methods optimize robot trajectories by ensuring that the robots’ time-averaged spatial distribution aligns with the spatial distribution of environmental information. While these methods promote effective exploration provided a target distribution, they often fail to account for unknown prior distributions of the environment. To overcome this limitation, we propose an adaptive coverage strategy that utilizes real-time feedback from an environmental model to adjust robot sampling behavior in response to unknown conditions. Our approach enhances traditional ergodic trajectory optimization by constructing a target spatial information distribution based on parametric models of the environment, which are updated online. This strategy assumes that the environment is either static or changes slowly compared to the robot's motion. Our framework allows robots to dynamically prioritize regions of high interest, improving coverage efficiency, synthesizing effective control policies for individual agents, and optimizing resource use in settings with unknown prior distributions. We validate our approach through simulations, demonstrating its effectiveness in enhancing coverage and resource allocation.

Keywords: Fluid flow systems, Time-varying systems, Lyapunov methods
Abstract: This work analyzes accelerating and decelerating wall-driven flows by quantifying the upper bound of transient energy growth using a Lyapunov-type approach. By formulating the linearized Navier-Stokes equations as a linear time-varying system and constructing a time-dependent Lyapunov function, we obtain an upper bound on transient energy growth by solving linear matrix inequalities. This Lyapunov method can obtain the upper bound of transient energy growth that closely matches transient growth computed via the singular value decomposition of the state-transition matrix of linear time-varying systems. Our analysis captures that decelerating base flows exhibit significantly larger transient growth compared with accelerating flows. Our Lyapunov method offers the advantages of providing a certificate of uniform stability and an invariant set to bound the solution trajectory.

Keywords: Fluid flow systems, Time-varying systems, Lyapunov methods
Abstract: This work demonstrates that the Lyapunov method can effectively identify the growth rate of a linear time-periodic system describing cold fresh water on top of hot salty water with a periodically time-varying background shear flow. We employ a time-dependent weighting matrix to construct a Lyapunov function candidate, and the resulting linear matrix inequalities are discretized in time using the forward Euler method. As the number of temporal discretization points increases, the growth rate predicted from the Lyapunov method or the Floquet theory will converge to the same value as that obtained from numerical simulations. Additionally, the Lyapunov method is used to analyze the most dangerous disturbance, and we also compare computational resource usage for the Lyapunov method, numerical simulations, and the Floquet theory.

Keywords: Switched systems, Stability of linear systems, Time-varying systems
Abstract: A novel, computationally-efficient dwell time estimation algorithm for linear, time-varying switched systems with variable state dimension is proposed. The algorithm builds upon an existing method for fixed-dimension, linear, time-invariant (LTI) systems based on a Jordan form decomposition. This existing method is extended to variable-dimension, LTI switched systems then to variable-dimension, linear, time-varying (LTV) switched systems. The latter yields a time-varying dwell time estimate from the frozen-in-time subsystems within the overall LTV system. A dwell time estimate can be obtained using the maximum or mean over the time interval. Using the Floquet transformation, the LTI method is extended directly to periodic LTV systems. A numerical example of a variable-dimension, periodic LTV switched system illustrates the computation and interpretation of the proposed dwell time estimation method and compares it with other approaches.

Keywords: Time-varying systems, Uncertain systems, Modeling
Abstract: The work presented here investigates the application of polynomial chaos expansion toward input shaper design in order to maintain robustness in dynamical systems subject to uncertainty. Furthermore, this work intends to specifically address time-varying uncertainty by employing intrusive polynomial chaos expansion. The methodology presented is validated through numerical simulation of intrusive polynomial chaos expansion formulation applied to spring mass system experiencing time-varying uncertainty in the spring stiffness. The system also evaluates non-robust and robust input shapers through the framework in order to identify designs that minimize residual energy. Results indicate that vibration mitigation is achieved at a similar accuracy, yet at higher efficiency compared to a Monte Carlo framework.

Keywords: Optimization, Optimization algorithms, Decentralized control
Abstract: Decentralized optimization has recently gained attention due to its broad applications in modern high performance computing. In such settings, agents communicate over flexible and cost-effective networks that can be dynamically reconfigured. To leverage this flexibility, recent work has studied time-varying, sparse, and even disconnected network sequences in decentralized algorithms. We introduce DCATS, a dual decentralized coordinate ascent method designed for these communication structures. By accommodating sparse, time-varying graphs, DCATS achieves lower per-iteration communication cost than existing dual decentralized methods. We further provide a novel analysis capturing the joint effect of the communication graphs and establish linear convergence guarantees under the L-smooth and strongly convex setup. Numerical experiments support our theoretical findings, demonstrating the superiority of DCATS over baselines.

Keywords: Reinforcement learning, Learning, Large-scale systems
Abstract: We investigate a decentralized reinforcement learning problem involving multiple agents that interact with the same Markov Decision Process (MDP). The agents can exchange information over a network to collectively learn the optimal state-action value function. For this setting, we introduce a novel epoch-based distributed Q-learning algorithm called VRDQ, where within each epoch, agents locally estimate the Bellman optimality operator and diffuse information using a consensus-based protocol. For both static and time-varying networks, we establish high-probability finite-time convergence rates for VRDQ that enjoy linear speedups from collaboration. Crucially, we prove that such speedups in sample-complexity require only tilde{O}(1) communication, substantially improving upon the communication costs in prior work.

Keywords: Large-scale systems, Control of networks, Fault tolerant systems
Abstract: Complex, interconnected cyber-physical systems (CPS) are increasingly prevalent in domains such as power systems. Cyber-resilient architectures have been proposed to recover compromised cyber components of CPS. Recent works have studied tuning the recovery times of such architectures to guarantee safety in single-system settings. Extending these designs to interconnected CPS is more challenging, since solutions must account for attacks on multiple subsystems that can occur in any order and potentially infinite possible temporal overlap. This paper aims to address the aforementioned challenge by developing a scalable framework to assign resilient architectures and to inform the tuning of their recovery times. Our approach introduces a scalar index that quantifies the impact of each subsystem on safety under compromised input. These indices aggregate linearly across subsystems, enabling scalable analysis under arbitrary attack orderings and temporal overlaps. We establish a linear inequality relating each subsystem’s index and recovery time that guarantees safety and guides resilient architecture assignment. We also propose a segmentation-based approach to strengthen the previously derived conditions. We then present algorithms to compute the proposed indices and to find a cost-optimal architecture assignment with a safety guarantee. We validate the framework through a case study on temperature regulation in interconnected rooms under different attack scenarios.

Keywords: Large-scale systems, Distributed control, Constrained control
Abstract: This paper studies the safe control of very large multi-agent systems via a generalized framework that employs so-called Banach Control Barrier Functions (B-CBFs). Modeling a large swarm as probability distribution over a spatial domain, we show how B-CBFs can be used to appropriately capture a variety of macroscopic constraints that can integrate with large-scale swarm objectives. Leveraging this framework, we define stable and filtered gradient flows for large swarms, paying special attention to optimal transport algorithms. Further, we show how to derive agent-level, microscopical algorithms that are consistent with macroscopic counterparts in the large-scale limit. We then identify conditions for which a group of agents can compute a distributed solution that only requires local information from other agents within a communication range. Finally, we showcase the theoretical results over swarm systems in the simulations section.

Keywords: Large-scale systems, Stability of nonlinear systems, Network analysis and control
Abstract: This paper establishes sufficient conditions for achieving scalable Incremental Input-to-State Stability in large-scale multi-agent systems with arbitrary topologies. These results enable local tests at the agent level to ensure global stability properties. Using a trajectory-based approach, we derive explicit bounds on the deviations between system trajectories, accounting for differences in initial conditions and external inputs. The theoretical results are illustrated with an example that includes a plug-and-play scenario.

Keywords: Control of networks, Optimal control, Large-scale systems
Abstract: We exploit the convergence of graph sequences to some limit object, a graphon, to generate scalable initial estimates for optimal control of large-scale networks. We derive a state error bound under the assumptions of converging graph sequences and initial state functions and, given this bound, we propose an iterative scale-up approach to accelerate the time required for numerical solvers to reach the optimum. The idea is to scale up the optimal control solution for a network dynamical system and use it to initialize the optimal control problem for a structurally similar but larger network system. The numerical results demonstrate a significant reduction of the number of optimization iterations and the optimization time for a nonlinear large-scale coupled network.

Keywords: Decentralized control, Mean field games, Large-scale systems
Abstract: In this article, we employ an input-output approach to expand the study of cooperative multi-agent control and optimization problems characterized by mean-field interactions that admit decentralized and selfish solutions. The setting involves n independent agents that interact solely through a shared cost function, which penalizes deviations of each agent from the group’s average collective behavior. Building on our earlier results established for homogeneous agents, we extend the framework to nonidentical agents and show that, under a diagonal dominant interaction of the collective dynamics, with bounded local open-loop dynamics, the optimal controller for mathcal{H}_infty and mathcal{H}_2 norm minimization remain decentralized and selfish in the limit as the number of agents n grows to infinity.

Keywords: Optimization algorithms, Large-scale systems, Network analysis and control
Abstract: We present a GPU-accelerated proximal message passing algorithm for large-scale network utility maximization (NUM). NUM is a fundamental problem in resource allocation, where resources are allocated across various streams in a network to maximize total utility while respecting link capacity constraints. Our method, a variant of ADMM, requires only sparse matrix–vector multiplies with the link–route matrix and element-wise proximal operator evaluations, enabling fully parallel updates across streams and links. It also supports heterogeneous utility types, including logarithmic utilities common in NUM, and does not assume strict concavity. We implement our method in PyTorch and demonstrate its performance on problems with tens of millions of variables and constraints, achieving 4x to 20x speedups over existing CPU and GPU solvers and solving problem sizes that exhaust the memory of baseline methods. We also show that our algorithm is robust to congestion and link-capacity degradation. Finally, using a time-expanded transit seat allocation case study, we illustrate how our approach yields interpretable allocations in realistic networks.

Keywords: Machine learning, Game theory, Stochastic optimal control
Abstract: We show that common estimators for Nemytskii Mean-Field-Type Games including kernel density estimators, multi-level Monte Carlo schemes, neural networks, transformers, and truncated Wiener Chaos Expansions are structurally biased. Unlike variance, this bias arises from bandwidth, architectural, or truncation constraints and persists even with infinite samples. As a result, evaluating payoffs and seeking equilibria are flawed: equilibria of biased approximations do not approximate the true game. This reflects the non-commutativity of optimization and approximation. These findings call for unbiased or debiased estimation frameworks.

Keywords: Machine learning, Kalman filtering, Filtering
Abstract: The Kalman filter is optimal for state estimation in linear dynamic systems. For nonlinear systems, the main challenge is propagating uncertainty through the state transition and observation functions. Neural network surrogate models admit accurate uncertainty propagation using recent analytic formulas for the mean and covariance of deep neural networks with Gaussian input. We harness this technology for Kalman filtering by combining uncertainty propagation with a neural network implementation of a certain functional semigroup. We demonstrate this approach on the stochastic Lorenz system with a sampling time 100 times longer than in prior work. Relative to the Extended and Unscented filters and smoothers, our method achieves a sevenfold reduction in estimation error and superior coverage of the true state by its confidence regions. A full version of this paper, containing proofs, algorithms, and other details, is available at https://arxiv.org/abs/2511.09016.

Keywords: Machine learning, Neural networks, Optimization
Abstract: This work investigates goal-reaching control synthesis for neural network control systems. A backward reachability framework is developed based on constrained zonotopes, in which the graph set of a ReLU-activated feedforward neural network is encoded as a finite union of constrained zonotopes. Using this representation, under-approximations of backward reachable sets are computed for systems with nonlinear plant models, ensuring the feasibility of the goal-reaching task. Control sequences are then synthesized through an optimization procedure that exploits the under-approximated set. A numerical example demonstrates the effectiveness of the proposed approach.

Keywords: Machine learning, Neural networks, Optimization
Abstract: This work addresses the safe control synthesis problem for neural network control systems subject to bounded unknown disturbances and known exogenous inputs. A forward reachability analysis method is developed to over-approximate the system's forward reachable sets using constrained zonotopes, where the control sequence appears linearly in both the zonotope center and the right-hand side of the associated equality constraints. Based on these over-approximations, a quadratically constrained program and its convexification are formulated to synthesize control sequences that guarantee safety. A numerical example demonstrates the effectiveness of the proposed approach.

Keywords: Machine learning, Nonlinear systems identification, Feedback linearization
Abstract: This paper presents a control-oriented learning framework for physical reservoir computing (pRC) using underactuated nonlinear mass-spring-damper network, as a reservoir, for emulating arbitrary nonlinear systems with stable limit cycles. For a given nonlinear dynamics, partial feedback equivalent systems were derived to linearize the active subsystem and exploit the passive nonlinear subsystem of the physical reservoir. The stability of the linearized active subsystem and the zero dynamics were enforced for both linear and nonlinear readout functions. Required stability conditions were determined, leading to a set of constraints associated with Lyapunov stability for training readout weights. The effectiveness of the Lyapunov stability constraint-based learning framework was demonstrated in emulating the Van der Pol (VDP) and Quadratic limit cycle (QLC) dynamics without requiring washout time or fully known training data, and it was compared against the standard open-loop linear regression training approach. A closed-loop dynamics-informed learning process was established, which simulates the dynamics of a physical reservoir at a 1 kHz rate and trains the readout weights using the "teacher forcing" method, subject to stability constraints at a 1 Hz rate. Finally, the trained pRC was simulated, where both VDP and QLC oscillators were regenerated.

Keywords: Computer-aided control design, Evolutionary computing, Reinforcement learning
Abstract: Large Language models (LLMs) have shown promise as generators of symbolic control policies, producing interpretable program-like representations through iterative search. However, these models are not capable of separating the functional structure of a policy from the numerical values it is parametrized by, thus making the search process slow and inefficient. We propose a hybrid approach that decouples structural synthesis from parameter optimization by introducing an additional optimization layer for local parameter search. In our method, the numerical parameters of LLM-generated programs are extracted and optimized numerically to maximize task performance. With this integration, an LLM iterates over the functional structure of programs, while a separate optimization loop is used to find a locally optimal set of parameters accompanying candidate programs. We evaluate our method on a set of control tasks, showing that it achieves higher returns and improved sample efficiency compared to purely LLM-guided search. We show that combining symbolic program synthesis with numerical optimization yields interpretable yet high-performing policies, bridging the gap between language-model-guided design and classical control tuning. Our code is available at https://sites.google.com/berkeley.edu/colmo.

Keywords: Kalman filtering, Estimation, Observers for Linear systems
Abstract: This paper revisits the problem of guaranteeing stability of the Kalman predictor for discrete-time, linear, time-varying systems. The main result is a sufficient condition for Lyapunov stability and convergence of the state-estimation error to zero under weaker conditions than assumed in prior work.

Keywords: Kalman filtering, Filtering, Estimation
Abstract: Extended Object Tracking (EOT) seeks to jointly estimate object kinematics and shape from noisy point cloud data. Gaussian Process Implicit Surfaces (GPIS) enable the probabilistic representation of complex geometries, but existing GPIS-based tracking methods rely on Particle Filters (PF), which are too costly for real-time deployment. We propose a more efficient alternative using linearization-based state estimation within the GPIS framework. The measurement model is formulated with GPIS predictive moments, and an Extended Kalman Filter with analytical Jacobians avoids particle-wise evaluations. Input uncertainty is handled via first-order Taylor expansion. Monte Carlo simulations on varied geometries that show our method achieves comparable accuracy to PF-based methods, improves shape reconstruction, and greatly reduces computational cost, enabling real-time GPIS-based tracking for safety-critical applications.

Keywords: Distributed control, Kalman filtering, Decentralized control
Abstract: This paper studies the problem of distributed state estimation (DSE) over sensor networks on matrix Lie groups, which is crucial for applications where system states evolve on Lie groups rather than vector spaces. We propose a diffusionbased distributed Unscented Kalman Filter on Lie groups using the inverse covariance intersection (DFICI-UKF) method to address target tracking in 3D environments. There are two major distinctions between this work and existing results. First, unlike existing diffusion-based DSE methods confined to vector spaces, our approach extends such framework to Lie group settings, enabling local estimates to be fused with intermediate information from neighboring agents on Lie groups. Second, to handle the unknown correlations across local estimates, we extend the ICI fusion strategy to matrix Lie groups for the first time and integrate it into the diffusion algorithm. We theoretically prove that the estimation error of the proposed method is bounded. The algorithm is fully distributed, robust against intermittent measurements, and adaptable to timevarying communication topologies. The effectiveness of the proposed method is validated through extensive Monte-Carlo simulations.

Keywords: Kalman filtering, Nonlinear systems identification, Estimation
Abstract: Nonlinear extensions of the Kalman filter (KF), such as the extended Kalman filter (EKF) and the unscented Kalman filter (UKF), are indispensable for state estimation in complex dynamical systems, yet the conditions for a nonlinear KF to provide robust and accurate estimations remain poorly understood. This work proposes a theoretical framework that identifies the causes of failure and success in certain nonlinear KFs and establishes guidelines for their improvement. Central to our framework is the concept of covariance compensation: the deviation between the covariance predicted by a nonlinear KF and that of the EKF. With this definition and detailed theoretical analysis, we derive three design guidelines for nonlinear KFs: (i) invariance under orthogonal transformations, (ii) sufficient covariance compensation beyond the EKF baseline, and (iii) selection of compensation magnitude that favors underconfidence. Both theoretical analysis and empirical validation confirm that adherence to these principles significantly improves estimation accuracy, whereas fixed parameter choices commonly adopted in the literature are often suboptimal. The codes and the proofs for all the theorems in this paper are available at https://github.com/Shida-Jiang/Guidelines-for-Nonlinear-Kal man-Filters

Keywords: Kalman filtering, Observers for nonlinear systems
Abstract: The Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) are widely used for nonlinear state estimation. The EKF employs first-order linearizations, while the UKF uses the unscented transformation to more accurately capture higher-order moments. Despite their effectiveness, both filters can suffer from limitations such as linearization errors, biased estimates, and potential instability when subject to significant nonlinearities. This paper shows that the performance of the EKF can be improved by compensating for the effects of the higher-order terms that are typically neglected during the linearization of the measurement equation in the EKF derivation through an adaptive mechanism. Simulation results over several Monte Carlo runs across two nonlinear systems show that our proposed Adaptive Extended Kalman Filter (AEKF) can adaptively learn and compensate for these uncertainties, improving estimation accuracy and consistently outperforming both the traditional EKF and the UKF.

Keywords: Kalman filtering, Sensor networks, Neural networks
Abstract: We propose a Neural-Enhanced Distributed Kalman Filter (NDKF) for multi-sensor state estimation in nonlinear systems. Unlike traditional Kalman filters that rely on explicit analytical models and assume centralized fusion, NDKF leverages neural networks to replace analytical process and measurement models with learned mappings while each node performs local prediction and update steps and exchanges only compact posterior summaries with its neighbors. This distributed design reduces communication overhead and avoids a central fusion bottleneck. We provide sufficient mean-square stability conditions under bounded Jacobians and well-conditioned innovations, together with practically checkable proxies such as Jacobian norm control and innovation monitoring. We also discuss consistency under learned-model mismatch, including covariance inflation and covariance-intersection fusion when cross-correlations are uncertain. Simulations on a 2D nonlinear system with four partially observing nodes show that NDKF outperforms a distributed EKF baseline under model mismatch and yields improved estimation accuracy with modest communication requirements.

Keywords: Optimization, Optimal control, Neural networks
Abstract: In this work we study the convergence of gradient methods for nonconvex optimization problems -- specifically the effect of the problem formulation to the convergence behavior of the solution of a gradient flow. We show through a simple example that, surprisingly, the gradient flow solution can be exponentially or asymptotically convergent, depending on how the problem is formulated. We then deepen the analysis and show that a policy optimization strategy for the continuous-time linear quadratic regulator (LQR) (which is known to present only asymptotic convergence globally) presents almost global exponential convergence if the problem is overparameterized through a linear feed-forward neural network (LFFNN). We prove this qualitative improvement always happens for a simplified version of the LQR problem and derive explicit convergence rates for the gradient flow. Finally, we show that both the qualitative improvement and the quantitative rate gains persist in the general LQR through numerical simulations.

Keywords: Optimization, Stability of nonlinear systems, Stochastic systems
Abstract: This paper studies gradient dynamics subject to additive stochastic noise, which may arise from sources such as stochastic gradient estimation, measurement noise, or stochastic sampling errors. To analyze the robustness of such stochastic gradient systems, the concept of small-covariance noise-to-state stability (NSS) is introduced, along with a Lyapunov-based characterization. Furthermore, the classical Polyak–Łojasiewicz (PL) condition on the objective function is generalized to the mathcal{K}-PL condition via comparison functions, thereby extending its applicability to a broader class of optimization problems. It is shown that the stochastic gradient dynamics exhibit small-covariance NSS if the objective function satisfies the mathcal{K}-PL condition and possesses a globally Lipschitz continuous gradient. This result implies that the trajectories of stochastic gradient dynamics converge to a neighborhood of the optimum with high probability, with the size of the neighborhood determined by the noise covariance. Moreover, if the mathcal{K}-PL condition is strengthened to a mathcal{K}_infty-PL condition, the dynamics are NSS; whereas if it is weakened to a general positive-definite-PL condition, the dynamics exhibit integral NSS. The results further extend to objectives without globally Lipschitz gradients through appropriate step-size tuning. The proposed framework is further applied to the robustness analysis of policy optimization for the linear quadratic regulator (LQR) and logistic regression.

Keywords: Machine learning, Predictive control for nonlinear systems, Chemical process control
Abstract: Traditional model predictive control (MPC) relies on accurate mechanistic models, which are often costly or impractical to derive for complex systems. Machine learning-based MPC (ML-MPC) provides an attractive alternative by learning predictive models directly from process data. However, its adoption in industrial practice is hindered by concerns over data privacy. To address this challenge, we propose an encrypted RNN-based MPC (ERNN-MPC) that enables privacy-preserving training and efficient deployment of ML models. In this design, recurrent neural networks (RNNs) are trained over homomorphically encrypted data to ensure data confidentiality throughout the training stage. The encrypted trained RNN is then returned to the plant, where it is executed in a plaintext inference mode and integrated into the MPC framework. Finally, by embedding ERNN into a Lyapunov-based MPC formulation, we establish closed-loop stability guarantees for the proposed control framework. Validation on a chemical reactor demonstrates the effectiveness of the proposed approach.

Keywords: Hybrid systems, Machine learning
Abstract: Reservoir computing is a machine learning framework based on using the dynamics of a fixed high-dimensional nonlinear system (a reservoir) to predict spatiotemporal data or control a dynamical system. The expressiveness of the reservoir then allows behavior to be predicted by training only a linear readout vector, rather than training a higher number of vectors within the reservoir, resulting in fast and low-cost training. This framework has been shown to be successful using both continuous- and discrete-time systems. However, it has not been considered within the context of hybrid dynamical systems, which contain both continuous and discrete dynamics. Hybrid systems are able to exhibit a rich range of behaviors, even in low-dimensions, making them amenable to use as a reservoir. In this work we examine the theoretical constraints for using hybrid systems in reservoir computing. The main constraint for success in reservoir computing is the echo state property, which is based on the asymptotic behavior of the reservoir. We first provide a careful formulation of requirements for using a hybrid system as a reservoir before providing conditions such that the echo state property holds for different forms of hybrid reservoirs. We conclude with simulations illustrating the use of a hybrid reservoir satisfying the echo state property for prediction of a hybrid model of a bouncing ball.

Keywords: Stability of nonlinear systems, Optimization, Lyapunov methods
Abstract: Recent work has shown that evaluating a nonlinear dynamical system (i.e., computing a trajectory from an initial state) can be reformulated as an optimization problem, where each step can be parallelized. This perspective enables dramatic speed-ups on parallel hardware such as GPUs. In this work, we deepen this connection by showing that evaluating a contracting dynamical system is equivalent to optimizing a well-behaved loss function that satisfies the Polyak–Lojasiewicz condition (also known as gradient dominance), where the PL constant is monotone in the contraction rate and the conditioning of the contraction metric. Building on this insight, we further exploit the robustness of contracting systems to design memory-efficient alternatives to existing algorithms such as DEER and DeepPCR. We also characterize these latter two algorithms as predictor-corrector algorithms, connecting them to an existing literature. Finally, we experimentally validate our claims, offering practical guidance on when nonlinear dynamical systems can be efficiently parallelized, and highlighting contractivity as a key design principle for parallelizable models with provable theoretical guarantees on training/learning.

Keywords: Machine learning, LMIs, Optimization algorithms
Abstract: Classical data-driven methods can be conceptualized as mappings from data distributions to decisions. However, in practice, decisions can influence the data distributions themselves. One of the common methods for handling unknown decision-dependent distribution shift is repeated optimization. In this paper, we model repeated optimization as a discrete-time feedback interconnection system. Our framework enables convergence analysis base on dissipation inequalities and integral quadratic constraints, which provides a novel method to show convergence under unknown decision-dependent distribution shift. We bound the suboptimality when using repeated gradient descent and ignoring the distribution shift when taking gradient steps. Additionally, our framework provides a method to bound the distance between performatively stable points and performatively optimal points.

Keywords: Biomolecular systems, Cellular dynamics, Systems biology
Abstract: Biological neural networks (BioNNs) aim to replicate neural computation using living cells as distributed processing units. We introduce a population-scale BioNN perceptron architecture that embeds control-theoretic regulation into computation itself. Sender populations encode input weights through paradoxical population control, a feedback strategy that stabilizes population size and ensures robust, tunable signal outputs. A shared receiver population integrates these signals, applying molecular summation and sequestration-based thresholding to implement a perceptron decision rule. Using analytic control models, we derive explicit activation conditions that map biochemical parameters to classification performance, and validate these results through simulations of single- and multi-sender activation. By showing how feedback regulation can serve simultaneously as a stabilizing mechanism and a computational primitive, this framework establishes a foundation for robust, interpretable, and scalable BioNNs, opening the door to higher-dimensional classification in microbial consortia.

Keywords: Biomolecular systems, Feedback linearization, Biotechnology
Abstract: Synthetic biology is a promising field that is currently limited by the lack of consideration for biological phenomena, such as resource competition, during the genetic controller design phase, resulting in unpredictable experimental results. We aim to address this by developing a resource-aware model of a general two-gene network, using linearisation to characterise the impact of resource competition on controller performance, specifically reference tracking, transient response and disturbance rejection. We apply our analysis to a typical motif used to enact negative feedback and demonstrate how resource competition can be exploited to improve robustness. This work demonstrates that, despite being typically viewed as undesirable, resource competition can be used to enhance controller performance.

Keywords: Biomolecular systems, Genetic regulatory systems, Systems biology
Abstract: Spatiotemporal self-organization is an emergent phenomenon governing cellular behavior across diverse living systems with key implications for biological functionality. Understanding the biomolecular mechanisms governing intercellular coordination would allow for the design, optimization, and prediction of these systems. Recent advances combine genetic engineering with PDE models to study the spatiotemporal dynamics of bacterial colonies, while engineered bacteria also function as bioreporters linking environmental detection to observable responses. We propose a design that extends prior frameworks by coupling genetic systems governing cell density–mediated motility to an external target substance, enabling its spatial encoding. Unlike earlier PDE models where motility depends only on cell density, our approach introduces indirect dual regulation in which the target substance modulates the quorum-sensing signal controlling density-dependent motility, converting spatial self-organization into an input–output mapping. We present a PDE model capturing spatiotemporal feedback signaling and introduce geometric metrics quantifying substance-induced pattern perturbations. Using this model, we show that the system encodes information about the concentration of a target substance through predictable spatiotemporal patterns that can be used to infer properties of that substance. The overall goal is to motivate new approaches for diagnosis and substance detection and highlight the importance of model-guided design for genetically programmable feedback control of spatiotemporal behaviors.

Keywords: Biomolecular systems, Machine learning, Nonlinear systems identification
Abstract: In several applications, including in synthetic biology, one often has input/output data on a system composed of many modules, and although the modules’ input/output functions and signals may be unknown, knowledge of the composition architecture can allow to significantly reduce the amount of training data required to learn the system’s input/output mapping. Learning the modules’ input/output functions is also necessary for designing new systems from different composition architectures. Here, we propose a modular learning framework, which incorporates prior knowledge of the system’s compositional structure to (a) identify the composing modules’ input/output functions from the system’s input/output data and (b) achieve this by using a reduced amount of data compared to what would be required without knowledge of the compositional structure. To achieve this, we introduce the notion of modular identifiability, which allows to recover modules’ input/output functions from a subset of system’s input/output data, and provide theoretical guarantees on a class of systems motivated by genetic circuits. We demonstrate the theory on computational studies showing that a neural network (NNET) that accounts for the compositional structure can learn the composing modules’ input/output functions and predict the system’s output on inputs outside of the training set distribution. By reducing the need for experimental data and allowing modules’ identification, this framework offers the potential to ease the design of synthetic biological circuits and of multi-module systems more generally.

Keywords: Biological systems, Genetic regulatory systems, Biomolecular systems
Abstract: We present a method to regulate the average global supercoiling state of cellular DNA. Our approach begins from a stochastic description of transcription-induced torsional stresses, where gene-to-gene variability creates heterogeneous local supercoiling inputs. Using a chemical master equation, we capture this heterogeneity and apply moment-based averaging to derive mean-field dynamics for the aggregate system. The resulting nonlinear model captures the feedback loop in which enzyme activity regulates DNA topology, DNA topology modulates transcriptional capacity, and transcriptional activity drives enzyme production. We prove the existence of a nontrivial equilibrium and perform a sensitivity analysis to identify key parameters governing supercoiling homeostasis. Model validation is provided by simulations that realize temperature as a saturated actuator using the thermally sensitive E. coli gyrase mutant nalA43, demonstrating regulation of supercoiling and its impact on growth.

Keywords: Randomized algorithms, Constrained control, Autonomous robots
Abstract: The Resource Constrained Shortest Path Problem (RCSPP) is defined as finding a minimum cost path between two poses while maintaining a path dependent resource consumption below a prescribed limit. The RRT* algorithm operates by generating random vertices and connecting them to each other to find the minimum cost path between two poses. In this work we propose a simple adaptation to RRT* denoted RC-RRT* that guarantees feasible paths with respect to the resource constraint. We prove that RC-RRT* is no longer probabilistically complete when applied to the RCSPP. Additionally we propose a Lagrangian Relaxation based RRT search denoted Lambda-RRT* which minimizes the relaxed cost given some relaxation parameter. We propose a method to update the relaxation parameter with each node added such that the minimum relaxed cost path converges to a feasible path. Finally we test the performance of both RC-RRT* and Lambda-RRT in a set of randomly generated planning scenarios.

Keywords: Sampled-data control, Constrained control
Abstract: In this paper, we present a resource-aware control framework that leverages dynamic control rate adaptation to provide both safety and real-time schedulability across multiple concurrent control applications. Safety is enforced through a discrete-time control barrier function with a novel formulation that explicitly accounts for variable sampling periods. Schedulability is ensured by dynamically bounding the minimum admissible period under a given scheduling policy. Both constraints are encoded in an optimization-based controller, which adapts each sampling period, enabling longer periods when safe, and faster ones near the safe set margins. Simulations with multiple mobile robots show that the method preserves safety while improving resource utilization compared to classic fixed-period control.

Keywords: Machine learning, Pattern recognition and classification, Computer-aided control design
Abstract: In response to the growing demand for ultra-low-power intelligence in edge control applications—ranging from smart furniture to autonomous robotics—we introduce {em HyperSNN}, a co-designed software–hardware framework that unites spiking neural networks with hyperdimensional computing. By substituting energy-hungry multiplies with lightweight additions and bitwise operations, HyperSNN slashes energy consumption while enhancing resilience to sensor noise and model perturbations. We demonstrate a low-power reconfigurable 22nm ASIC design and evaluate it on standard OpenAI Gym control tasks (CartPole, Acrobot, MountainCar, and LunarLander), achieving control performance on par with conventional machine-learning controllers but with nJ-level energy and up to 3.05times energy efficiency. This paper also discuss how HyperSNN’s efficiency and robustness pave the way for deploying advanced schemes—such as predictive-correction control—directly on resource-constrained edge devices.

Keywords: Autonomous systems, Multivehicle systems, Automotive control
Abstract: Multi-agent systems are integral tools for various domains such as robotics, cybersecurity, and autonomous vehicle planning. These types of systems often have constraints on the computational resources, leading to a need for efficient lightweight algorithms. Traditional decision making frameworks often assume ideal conditions, such as full observability and unlimited computational capacity, which do not align with real-world challenges. In this paper, we introduce a new algorithm that allows for reduced demand on computational resources without a large cost of other performance metrics. Agents will limit their observability to some attention radius, which intentionally allows them to ignore parts of the environment that might be unnecessary for action planning. By optimizing both the attention radius and decision-making, our approach enhances coordination and scalability in uncertain environments. Through both theoretical analysis and empirical validation, we demonstrate the effectiveness of adaptive observation in improving system performance and maintaining robust decision-making strategies in resource-constrained systems.

Keywords: Information theory and control, Game theory, Control applications
Abstract: We study sequential information acquisition from a Gaussian source when attention is limited and the stage criterion may change over time. At each stage, the decision maker selects an information matrix that updates the posterior covariance. We first derive a posterior-domain reformulation of the stage problem and identify conditions under which an optimal action may be chosen to commute with the current prior covariance. Under a class of Schur-convex attention measures, the reduced scalar problem admits a water-filling solution and the optimizer is invariant across the corresponding family of unitarily invariant norms. This yields a norm-invariant stage rule when the realized stage criterion is revealed online and only per-stage attention constraints are present. We then formulate the finite-horizon problem with both stagewise and cumulative attention budgets as a dynamic program. The state, feasible action correspondence, and grid projection used in the numerical approximation are specified explicitly. Numerical examples illustrate the distinction between the norm-invariant regime and schedule-dependent regimes induced by non-unitarily-invariant criteria.

Keywords: Manufacturing systems, Sensor fusion, Embedded systems
Abstract: In industrial applications such as semiconductor manufacturing, high-precision, high-speed positioning is critical for maintaining assembly throughput. These systems rely on fast and accurate sensor feedback, often achieved by fusing heterogeneous data to enable precise motion control. In our earlier work, we demonstrated multi-rate sensor fusion, combining rapid but less precise linear encoders with slower yet highly accurate vision-based detection. However, vision algorithms are computationally demanding, often exceeding available sampling periods and degrading closed-loop control performance. We address these challenges using the predictable multiprocessor CompSOC platform, which ensures deterministic execution and supports parallelization of vision algorithms. We systematically analyze trade-offs by exploring two design knobs: (i) vision parallelization, which reduces the sampling period while requiring additional hardware resources, and (ii) vision subsampling and range reduction, which reduce the sampling period but decrease vision sensor precision. Furthermore, we evaluate the effect of semiconductor die orientation on qualityof-control (QoC). By incorporating these analyses into Pareto trade-offs between control performance and computational resources, our approach provides practical insights for system designers. It helps optimize hardware resources and sensor strategies for high-precision, resource-constrained industrial applications, thereby achieving high QoC. We demonstrate this on a multisensor positioning system implemented on a predictable multiprocessor platform in a semiconductor motion stage case study.

Keywords: Optimal control, Neural networks, Machine learning
Abstract: Neural network approaches that parameterize value functions have succeeded in approximating high-dimensional optimal feedback controllers when the Hamiltonian admits explicit formulas. However, many practical problems, such as the space shuttle reentry problem and bicycle dynamics, among others, may involve implicit Hamiltonians that do not admit explicit formulas, limiting the applicability of existing methods. Rather than directly parameterizing controls, which does not leverage the Hamiltonian's underlying structure, we propose an end-to-end implicit deep learning approach that directly parameterizes the value function to learn optimal control laws. Our method enforces physical principles by ensuring trained networks adhere to the control laws by exploiting the fundamental relationship between the optimal control and the value function's gradient; this is a direct consequence of the connection between Pontryagin's Maximum Principle and dynamic programming. Using Jacobian-Free Backpropagation (JFB), we achieve efficient training despite temporal coupling in trajectory optimization. We show that JFB produces descent directions for the optimal control objective and experimentally demonstrate that our approach effectively learns high-dimensional feedback controllers across multiple scenarios involving implicit Hamiltonians, which existing methods cannot address.

Keywords: Optimal control, Optimization, Reinforcement learning
Abstract: We analyze the performance of policy gradient in multitask linear quadratic regulation (LQR), where the system and cost parameters differ across tasks. The main goal of multitask LQR is to find a controller with satisfactory performance on every task. Prior analyses on relevant contexts fail to capture closed-loop task similarities, resulting in conservative performance guarantees. To account for such similarities, we propose bisimulation-based measures of task heterogeneity. Our measures employ new bisimulation functions to bound the cost gradient distance between a pair of tasks in closed loop with a common stabilizing controller. Employing these measures, we derive suboptimality bounds for both the multitask optimal controller and the asymptotic policy gradient controller with respect to each of the tasks. We further provide conditions under which the policy gradient iterates remain stabilizing for every system. For multiple random sets of certain tasks, we observe that our bisimulation-based measures improve upon baseline measures of task heterogeneity dramatically.

Keywords: Optimal control, Predictive control for linear systems, Constrained control
Abstract: This letter focuses on a formulation of Model Predictive Control (MPC) with an optimal control problem (OCP) defined by hard input constraints and soft state and terminal set constraints. The soft constraints are accounted for as relaxed barrier function terms in the objective function. The proposed MPC is feasible for any state vector and, assuming the input constraint set is simple (e.g. a hyperrectangle), leads to anytime feasible formulations. A theoretical description of the MPC scheme is conducted. Among other results, asymptotic stability of the proposed MPC is proven and a region of attraction (RoA) estimate is derived. Moreover, stability guarantees when performing a limited number of optimization iterations are also derived. Numerical results showcase the benefit of considering the input constraints directly in the OCP instead of saturating the output of an unconstrained OCP with relaxed barrier functions, as was previously done in the literature.

Keywords: Optimal control, Predictive control for linear systems, Manufacturing systems
Abstract: This paper proposes a dynamic optimization framework that jointly optimizes production planning and waste acquisition in recycling-based Additive Manufacturing (AM) supply chains. While production planning and raw material acquisition are each well studied, existing approaches treat them independently—leaving production schedules decoupled from material sourcing decisions. The proposed model closes this gap by introducing a dual-source raw material structure, pairing limited customer-supplied waste with a higher-cost external supplier, and by treating waste pickup as a controllable decision variable rather than a passive process. The resulting mixed integer linear program is solved within a Model Predictive Control framework. An optimality analysis based on the Karush– Kuhn–Tucker conditions yields a reduced-cost signal that governs waste pickup through three economically interpretable regimes. Numerical results show that the framework eliminates service failures under material shortages and reduces total cost by more than 60% under surplus conditions.

Keywords: Optimal control, Variational methods, Linear systems
Abstract: Time-optimal control for triple integrator under full box constraints is a fundamental problem in the field of optimal control, which has been widely applied in the industry. However, scenarios involving asymmetric constraints, non-stationary boundary conditions, and active position constraints pose significant challenges. This paper provides a complete characterization of time-optimal switching surfaces for the problem, leading to novel insights into the geometric structure of the optimal control. The active condition of position constraints is derived, which is absent from the literature. An efficient algorithm is proposed, capable of planning time-optimal trajectories under asymmetric full constraints and arbitrary boundary states, with a 100% success rate. Computational time for each trajectory is within approximately 10μs, achieving a 5-order-of-magnitude reduction compared to optimization-based baselines.

Keywords: Optimal control, Stability of nonlinear systems
Abstract: This paper proposes a gradient-based control co-design method for nonlinear optimal regulator problems, where physical design parameters and feedback controllers are optimized simultaneously. The proposed method is based on Galerkin approximations of the Hamilton–Jacobi–Bellman equation in a policy iteration framework. The key idea is to evaluate closed-loop performance as the expected cost over a prescribed distribution of initial states, which enables sensitivity analysis and gradient-based updates of the design parameters, while the controller is improved through policy iteration. As a result, the proposed method overcomes restrictive structural assumptions such as system equivalence, thereby avoiding conservatism and allowing flexible incorporation of design-dependent costs. Moreover, closed-loop stability is ensured at every iteration of the co-design procedure by embedding a recursive admissibility verification that combines two complementary Lyapunov conditions. The effectiveness of the proposed method is demonstrated through an example of a load positioning system.

Keywords: Predictive control for linear systems, Robust control, Uncertain systems
Abstract: This paper proposes a robust tube-based model predictive control approach for spacecraft rendezvous subject to sector-bounded nonlinearities and bounded disturbances. Unlike existing methods that design the feedback controller and tube tightening parameters sequentially, we jointly optimize both through a convex Linear Matrix Inequality using static quadratic constraints. This eliminates the conservatism inherent in two-step design procedures, while maintaining computational tractability for real-time implementation. The approach is validated through a CubeSat docking simulation with tight operational constraints, showing around 70% fuel savings, 21% reduction in average computational time and smaller tube sizes compared to LQR-based fixed-gain methods.

Keywords: Predictive control for linear systems, Optimal control, Optimization algorithms
Abstract: In model predictive control (MPC), the choice of cost-weighting matrices and designing the Hessian matrix directly affects the trade-off between rapid state regulation and minimizing the control effort. However, traditional MPC in quadratic programming relies on fixed design matrices across the entire horizon, which can lead to suboptimal performance. This study presents a Riccati equation-based method for adjusting the design matrix within the MPC framework, which enhances real-time performance. We employ a penalized least-squares (PLS) approach to derive a quadratic cost function for a discrete-time linear system over a finite prediction horizon. Using the method of weighting and enforcing the equality constraint by introducing a large penalty parameter, we solve the constrained optimization problem and generate control inputs for forward-shifted horizons. This process yields a recursive PLS-based Riccati equation that updates the design matrix as a regularization term in each shift, forming the foundation of the regularized MPC (Re-MPC) algorithm. To accomplish this, we provide a convergence and stability analysis of the developed algorithm. Numerical analysis demonstrates its superiority over traditional methods by allowing Riccati equation-based adjustments.

Keywords: Predictive control for linear systems, Optimal control, Large-scale systems
Abstract: For large interconnected systems, centralized model predictive control (MPC) may be computationally intractable while distributed MPC may not achieve global optimum. To fill this gap, a new MPC framework, named asynchronous MPC, has been proposed in literature, where only a subset of subsystems is selected for optimization at each time step. For the remaining subsystems, the previously optimized control sequence is shifted and reused. Such an approach can balance computational load and control performance, and can be viewed as a framework in between centralized MPC and distributed MPC. However, prior work assumes each subsystem to be deterministic and does not consider the impacts of external disturbance. Moreover, feasibility and stability analysis for asynchronous MPC has not been studied. This paper addresses these limitations by (i) incorporating the prediction error introduced by external disturbance into reconfiguration policy, (ii) providing a stability condition for the closed-loop system, and (iii) establishing the recursive feasibility. Numerical results using battery cell-to-cell balancing control confirms the stability of the closed-loop system and demonstrates over 86% computation reduction.

Keywords: Predictive control for linear systems, Sampled-data control, Uncertain systems
Abstract: This letter presents an output-feedback tube-based model predictive control (MPC) framework for linear sampled-data control systems subject to external disturbances and non-convex constraints. The proposed approach rigorously incorporates inter-sample reachability analysis to account for the continuous-time evolution of system trajectories between discrete sampling instances and to ensure constraint satisfaction in the continuous time domain. The resulting continuous-time tube-based MPC scheme is demonstrated to ensure that trajectories remain within (potentially non-convex) safe sets throughout the continuous-time evolution.

Keywords: Predictive control for linear systems, Stochastic optimal control, Aerospace

Keywords: Predictive control for linear systems, Traffic control, Optimization
Abstract: Improving energy efficiency in the transportation sector is critical for achieving sustainable mobility, with eco-driving emerging as a key strategy. However, implementing effective eco-driving for connected and automated vehicles (CAVs) in mixed traffic presents a significant control challenge due to the heterogeneous, uncertain behavior of human-driven vehicles (HDVs). Data-enabled Predictive Control (DeePC) offers a promising model-free approach but is often hindered by a high computational burden, limiting its real-time feasibility. This paper introduces a novel Direct Data-driven Predictive Control (D3PC) framework to address this limitation. By reformulating the data-driven prediction mechanism, the D3PC significantly reduces computational complexity, making its computation time nearly invariant to historical data size. This computational efficiency directly enables the formulation of a sophisticated eco-driving controller that can solve the complex energy optimization problem in real time, even within diverse and stochastic mixed-traffic environments. Comprehensive simulations demonstrate that the D3PC is orders of magnitude faster than existing DeePC-based methods while achieving superior energy efficiency. Specifically, it reduces total platoon energy consumption by up to 10.71% compared to rule-based cruise control baselines and 3.80% compared to the original DeePC, confirming its effectiveness for real-time, energy-efficient control.

Keywords: Optimization, Robust control, Control software
Abstract: We present linrax, the first simplex based linear program (LP) solver in the JAX ecosystem. In many control algorithms, LPs are often automatically generated and solved online in the control loop. This motivates the design of linrax, which is especially suited for compilation into a large JAX-based pipeline as a subroutine. We discuss the challenges of implementing a general purpose LP solver under strict design requirements from JAX. Notably, we can solve general problems which may include dependent constraints—something not possible with existing JAX-compatible LP solvers that use first-order techniques and may fail to converge. We demonstrate linrax’s utility through several examples, including a robust control synthesis pipeline for nonlinear vehicles using automatic differentiation through a LP-based reachable set framework.

Keywords: Optimization algorithms, Predictive control for linear systems
Abstract: This work takes a first step towards updating branch-and-bound iterations online for mixed-integer quadratic model predictive control (MIQP MPC) while keeping empirical stability, i.e., stability in simulation, of the closed-loop control system. To be exact, the branch-and-bound iteration limits are adaptively updated after sufficiently many MPC control updates. The adaptive update law is a Hybrid Lyapunov like function and a branch-and-bound algorithm is proposed for clear analysis of the adaptive update law. Simulations verify the efficacy of the adaptive update law to reduce branch-and-bound iterations and maintain empirical stability of the feedback system for the controlled bouncing ball and minimum thrust spacecraft rendezvous problems.

Keywords: Optimization algorithms
Abstract: Compared with centralized optimization, distributed optimization is implemented within the multi-agent network, which improves computation efficiency and multi-agent collaboration under large-scale situations. This paper proposes a distributed convex optimization algorithm featuring a backtracking step size strategy without the prior knowledge of Lipschitz constant of the gradient of the objective function. Meanwhile, the convergence with the range of corresponding step size is given under the backtracking strategy. Finally, numerical examples are also given to validate the performance of the proposed algorithm.

Keywords: Optimization algorithms
Abstract: This paper presents a highly efficient Hybrid Sequential Linear Programming (SLP) Trust Region framework for convex, twice-continuously differentiable nonlinear programming (NLP) problems. By dynamically computing adaptive linearization error limits, the proposed method constructs strictly conservative, axis-aligned hyper-rectangular trust regions around interim solutions. Unlike standard algorithms relying on heuristic step sizes, this geometric approach guarantees strict monotonic feasibility while drastically reducing the size of the active constraint set. Crucially, by mapping single-constraint bottlenecks to a continuous fractional knapsack problem, the framework completely bypasses dense O(n3) matrix inversions, resolving the local subproblem analytically in expected O(n) time. Comprehensive numerical benchmarks on highly coupled Quadratically Constrained Quadratic Programs (QCQPs) demonstrate that the proposed architecture achieves speedups of over two orders of magnitude against state-of-the-art interior-point and active-set solvers without compromising optimal convergence.

Keywords: Electrical machine control, Optimization
Abstract: Finding optimal reference currents that minimize electric losses across the entire machine operating range is essential for electric vehicle drivetrains. Due to the non-convexity of the torque equation for electric machines, optimal reference current generation is a non-convex optimization problem which is difficult to solve analytically. Semidefinite relaxation is a well understood and effective technique primarily in the study of optimal power flow, but can also provide advantages which have not been studied as extensively in the field of motor control. As this research shows, the optimal efficiency reference generation problem has properties which allow the semidefinite relaxation to be applied, providing both theoretical advantages in terms of provable optimality as well as practical advantages in terms of solver speed. Utilizing a convex solver with the semidefinite programming formulation results in provably optimal reference currents while utilizing a constrained nonlinear optimization solver with the non-convex formulation cannot in general find the global optimum. The median solver time for the convex formulation also provided an 89% reduction relative to the non-convex solver time for a sample of 2500 points across the operating range of the wound rotor synchronous machine considered in this research.

Keywords: Optimization algorithms
Abstract: In most existing asynchronous methods, the stepsize depends on an upper bound on the delays and decreases as this bound increases. However, since the upper bound is usually unknown and large, the resulting stepsizes are not only difficult to determine in practice but also overly conservative, which leads to slow convergence. To solve this issue, we propose an adaptive stepsize strategy for a typical asynchronous optimization method---the Incremental Aggregated Gradient (IAG) method. Unlike existing methods in the literature, our stepsize does not rely on any delay information and is less conservative, which leads to easier stepsize determination and faster convergence. Under standard assumptions, we provide the convergence rate of IAG with the proposed stepsize. Numerical experiments demonstrate the superior performance of our stepsize over alternative stepsize strategies.

Keywords: Distributed parameter systems, Simulation, Optimal control
Abstract: This work investigates an optimal control problem for scheduling repairs for a basic reparable system described by coupled partial and ordinary differential equations. Our goal is to maximize the system availability with optimal repair strategy. This problem leads to a bilinear optimal control structure in non-smooth space, posing significant computational challenges. We analyze the model, address the first-order necessary conditions for optimality, and develop a numerical algorithm to solve them. The efficacy of our approach is demonstrated through numerical experiments, which validate the theoretical results and illustrate the practical implementation of the control strategy.

Keywords: Distributed parameter systems, Linear systems, Process Control
Abstract: A tubular reactor is studied, modeled by a two-dimensional parabolic PDE in cylindrical coordinates that incorporates radial and axial diffusion, axial convection, and reaction dynamics. Boundary actuation at the inlet and outlet measurement are employed to enable full-state feedback via pole placement and state reconstruction through a Luenberger observer. The effect of restricting the actuation domain to a smaller radial interval is analyzed to evaluate its influence on controller design. Numerical simulations reveal that gains sufficient for full inlet actuation may fail under restricted actuation, demonstrating the importance of a two-dimensional framework to accurately analyze and design controllers for restricted actuation scenarios in industrial applications.

Keywords: Distributed parameter systems, Fault accomodation, Process Control
Abstract: This work presents a framework for the development of actuator fault-tolerant control strategies for spatially distributed processes modeled by highly dissipative PDEs with limited sensor measurements and sensor-controller communication rate constraints. Using an approximate finite-dimensional model that captures the low-order dynamics of the infinite dimensional system, the framework brings together techniques from model-based control and state estimation, resulting in fault-tolerant control strategies that maintain closed-loop stability and reduce performance deterioration, while addressing the sensing and communication limitations. The proposed strategies are derived from analyses of the stability and performance properties of the finite-dimensional closed-loop system linking faults to a set of model and controller parameters that can be adjusted to ensure fault tolerance.

Keywords: Flexible structures, Differential-algebraic systems, Distributed parameter systems
Abstract: This paper presents a trajectory tracking strategy for controlling the shape of hyperelastic nonlinear flexible systems, formulated within the port-Hamiltonian Differential-Algebraic Equation (PH-DAE) framework. The control input is interpreted as a Lagrange multiplier enforcing trajectory constraints on the power-conjugated outputs of a finite element–based approximation of the model. By solving the multiplier analytically, a control law is obtained that guarantees exact trajectory enforcement. Alternatively, the control input is computed in discrete time by performing time integration while preserving the DAE structure, treating the input as an additional state. The proposed methods are validated by simulations on a von Kármán–based hyperelastic Timoshenko beam, using the Störmer–Verlet scheme for time integration, demonstrating the effectiveness of the approach.

Keywords: Stability of hybrid systems, Network analysis and control, Hybrid systems
Abstract: We consider a wave transmission network where two wave equations are joined at an interior interface and terminated at the free end by an inertia-generating boundary law (for example, a tip mass or an electrical load through a transducer). Sensing and actuation are an interface only. We compare two interface control designs: (A) a continuous coupling that enforces velocity continuity with mixed-order state feedback (velocity and angular velocity), and (B) a discontinuous coupling that permits a velocity jump and uses velocity-only state feedback with asymmetric gains, in the spirit of recent discontinuous interface paradigms ("Ozer-Walterman, 2025, 2026). For Design~A, we show that exponential decay fails, while strong stability holds conditionally. For Design~B, we prove uniform exponential decay of the total energy despite the destabilizing boundary inertia and non-collocation. The analysis employs interface multipliers and trace estimates to establish an observability-type integral bound, combined with an integral criterion to obtain an explicit decay rate with computable constants. Structure-preserving Finite Difference experiments support the theory, showing mesh-independent decay and clear spectral separation from the imaginary axis. The key takeaway is that a discontinuous, velocity-only interface feedback achieves exponential stabilization without boundary damping or higher-order terms, and its simplicity and locality make it promising for frame, tree, and mesh networks, plate or membrane analogues with line interfaces, and other wave-driven systems.

Keywords: Distributed parameter systems, Fluid flow systems, Automotive control
Abstract: We consider transport processes that are modeled by first order hyperbolic partial differential equations. Our goal is to find a full state feedback that makes a given reference profile locally asymptotically stable. To accomplish this we employ Linear Quadratic Regulation (LQR) with finite dimensional patch or point control actuation. We derive the Riccati partial differential equation whose solution is the kernel of the optimal cost. The optimal state feedback is also found. The derivation is accomplished by elementary techniques such as integration by parts and completing the square. We apply this theory to two examples that have appeared in the literature and that were solved by a modification of LQR. The first example deals with a model of a fixed-bed chemical reactor and the second example deals with traffic congestion on a stretch of freeway.