Keywords:
Abstract: This talk is by a control engineer who ended up working in agriculture, where the “factory” has no roof and the process runs on seasons, rain, and sunlight. The main problem Geco studies and solves is the control of weeds, which is usually considered more biological than control- theoretic, but it turns out there are plenty of dynamics involved.

We start at the beginning, which includes figuring out where a control approach actually makes sense. Agriculture comes with massive uncertainties just from being outdoors and biological. We will cover how the key problem was identified (not on our first try), why it is solvable and scalable, and how working closely with farmers and industry gave us the confidence to take each next step, and how we ended up finding places where multivariable and predictive control can make a real difference for farmers.

I will occasionally compare agriculture with other fields I have worked in, from paper machines to diesel engines, highlighting lessons learned from bringing control theory into the messy, real world.

Keywords: Control applications, Health and medicine
Abstract: In this paper, we present the development of a closed-loop drug delivery system for heart rate up-regulation in a novel ex vivo setup for functional evaluation of donor hearts. Such up-regulation is required to enable physiological testing aimed at supporting transplant decision-making. Clinical constraints prioritize safety over setpoint tracking performance. Accordingly, we develop a minimal dynamical model capturing the relevant response characteristics and use it to derive a closed-form upper bound on overshoot and tune a simple integrate-reset controller. We describe the real-time control system, which estimates heart rate from blood pressure measurements and regulates adrenaline infusion via a clinically certified pump. The controller design and modeling framework are discussed. Finally, the system is demonstrated in three ex vivo experiments on porcine hearts, showing closed-loop performance comparable to carefully executed manual drug titration. The controller is now used routinely in our ongoing evaluations.

Keywords: Health and medicine, Filtering, Identification
Abstract: Electrohysterography (EHG) enables non-invasive monitoring of uterine contractions but can be contaminated by electrocardiogram (ECG) artifacts. This work presents an ECG removal method using algebraic differentiators, a control-theoretic tool for model-free derivative estimation, that preserves signal shape outside the detected cardiac pulse locations. The differentiator parameters are designed to simultaneously suppress slow physiological artifacts and powerline interference. Cross-channel clustering distinguishes cardiac pulses from localized artifacts, enabling accurate pulse subtraction without auxiliary ECG references. Implemented as a causal FIR filter, the method is validated as a proof of concept on multichannel EHG recordings from one female and one male healthy volunteer and compared to the template subtraction method.

Keywords: Nonlinear systems, Control applications, Actuators
Abstract: This paper considers the noncontact manipulation of a magnetic particle inside a planar workspace encircled by an array of 3 electromagnets, the minimum number that allows for unconstrained planar manipulation of the particle. Since the size, weight, and cost of a magnetic manipulator directly depend on the number of its magnets, minimizing this number is highly desirable. Yet, magnetic manipulation with a minimum number of magnets exhibits certain limitations, which are investigated in this paper. It is shown both analytically and by experiment that while planar magnetic manipulation is feasible by an array of 3 electromagnets, it is less effective than what can be achieved by a larger number of magnets in terms of energy efficiency and the size of region in which magnetic manipulation is possible.

Keywords: Iterative learning control, Optimization, Manufacturing systems
Abstract: The thermal history of the melt pool in laser powder bed fusion (LPBF) additive manufacturing processes governs the solidification microstructure and the mechanical properties of the resulting 3D-printed parts. Dual-laser systems offer additional degrees of freedom to control the cooling profile by reheating material behind the melt pool, but calibrating process parameters is challenging due to the complex physics of the process. We present an optimization-based iterative learning controller that determines optimal power, velocity, and offset settings by judiciously combining simulations and experiments: the model supplies search directions while feedback obtained from experiments on the real plant corrects for parameter errors, enabling convergence despite model inaccuracies. The approach is validated in simulation using a high-fidelity thermal model as a plant surrogate, with deliberate mismatches in absorption coefficient, latent heat treatment, and powder-bed effective conductivity between plant and model. Results show that the controller drives the plant cost down by over an order of magnitude and reaches a tight band of low-cost solutions across seeds, while model-only feedforward optimization stalls at a substantially higher plant cost despite appearing to converge on the surrogate.

Keywords: Adaptive control, Nonlinear systems, Robotics
Abstract: This paper proposes an immersion and invariance (I&I) adaptive hybrid force-motion controller (IHFMC) for a two-degree-of-freedom (DoF) robot manipulator in contact with a rigid planar surface with unknown orientation. Based on the constrained dynamics, we present the classical hybrid force-motion control (HFMC) that decouples motion and force subspaces. In order to account for uncertainty in the orientation of the surface, we develop an I&I adaptive approach. The initial closed-loop system is inspired by the non-adaptive HFMC depending on estimated parameters. This closed-loop is immersed into a target dynamics that evolves on an invariant manifold designed to be attractive. The proposed controller ensures global exponential convergence of the position and force tracking errors as well as the parameter estimation error under a persistency of excitation (PE) condition. Simulations on a 2-DoF arm validate the proposed method, and a comparison to the non-adaptive HFMC is given.

Keywords: Game theory, Stochastic/uncertain systems, Automotive applications
Abstract: This paper investigates approximate strategies for stochastic mean-field incentive Stackelberg games with time delays. In this framework, a leader designs an incentive mechanism to steer noncooperative followers-who act according to a Nash equilibrium-toward the team-optimal solution. Compared with existing studies, we establish new sufficient solvability conditions via a coupled high-dimensional Lyapunov-type equations (CHLEs). To overcome the computational intractability that arises as the population size grows, we develop a reduced-order approach exploiting the asymptotic structure of the CHLEs. Based on this structure, we construct a population-size-independent approximate strategy and we quantify its near-optimality in terms of the associated cost function. The proposed framework is validated through numerical simulations of a connected and automated vehicle (CAV) platoon under stochastic disturbances, sensing noise, network latency, and actuation delay. The results confirm that the reduced-order strategy accurately captures the dominant closed-loop behavior of the full-order solution, demonstrating both scalability and practical effectiveness of the proposed method.

Keywords: Game theory, Stochastic/uncertain systems, Predictive control
Abstract: This paper presents an efficient computational framework for the nonlinear receding-horizon Nash differential game with stochastic parameter uncertainty. Unlike conventional Monte Carlo methods, which incur high computational and storage costs, the proposed approach employs polynomial chaos expansion to reformulate the stochastic Nash differential game as a deterministic problem in an expanded state space. This transformation allows optimal control strategies and statistical moments to be computed by solving a single deterministic problem at each time instant, thereby eliminating the need for a large number of simulations. Numerical experiments using a tank network system show that the proposed method attains statistical accuracy comparable to that of Monte Carlo simulations while achieving computation times compatible with real-time implementation.

Keywords: Intelligent systems, Stochastic/uncertain systems, Planning
Abstract: Mixed human-AI teams often operate with limited explicit communication under task uncertainty, so effective coordination requires quickly identifying which latent task or environment model best explains the observed interaction. This paper develops implicit task-hypothesis inference, enabling an AI agent to maintain a posterior over a finite set of candidate task models using its own experience and passively observed human state trajectories. We model collaboration as a task-uncertain cooperative Markov decision process (MDP) and derive a fully recursive Bayesian update that combines the agent's experience with human data. Unobserved human actions are modeled and integrated using a bounded-rational, cooperative policy model. The resulting posterior can be used for a wide range of downstream tasks, including online decision rules that either act under the most likely hypothesis (maximum a posteriori) or account for uncertainty by weighting decisions across hypotheses according to the posterior. We analyze when human trajectories are informative by studying their sensitivity to human decision reliability and identifying conditions under which the human likelihood provides limited additional discrimination among hypotheses. Numerical results on grid-world rescue in maze environments with hidden environmental structures demonstrate faster hypothesis identification and higher cooperative returns than competing inference methods.

Keywords: Stochastic/uncertain systems, Learning
Abstract: Learning-based structural vibration mitigation must remain reliable under excitation uncertainty whose severity, spectral content, and temporal structure can deviate from nominal training conditions. Standard single-agent reinforcement learning (RL) typically optimizes expected return under a fixed disturbance model, which can yield policies that perform well in distribution yet exhibit brittle behavior and elevated tail risk under disturbance shift or strategically adverse realizations. This paper studies reliability-aware policy learning through a game-theoretic uncertainty abstraction: we formulate vibration mitigation as a two-player zero-sum Markov game between a controller and an admissible excitation agent, and train policies via tabular minimax-Q learning with a state-wise matrix-game solve. Using an earthquake-motivated single-degree-of-freedom base-excitation structure, we compare uncontrolled and PD baselines, single-agent Q-learning trained under stochastic excitation models, and minimax-Q learning trained against bounded admissible excitation actions. We evaluate policies under (i) a fixed learned minimax excitation policy and (ii) controller-specific worst-case excitation policies within the same admissible disturbance set, reporting both average response metrics and reliability-relevant measures. The results show that minimax training produces controllers that are harder to exploit and yield improved worst-case reliability profiles within the admissible disturbance set.

Keywords: Stochastic/uncertain systems, Planning
Abstract: We present a chance-constrained model predictive control (MPC) framework under Gaussian mixture model (GMM) uncertainty. Specifically, we consider the uncertainty that arises from predicting future behaviors of moving obstacles, which may exhibit multiple modes (for example, turning left or right). To address the multi-modal uncertainty distribution, we propose three MPC formulations: nominal chance-constrained planning, robust chance-constrained planning, and contingency planning. We prove that closed-loop trajectories generated by the three planners are safe. The approaches differ in conservativeness and performance guarantees. In particular, the robust chance-constrained planner is recursively feasible under certain assumptions on the propagation of prediction uncertainty. On the other hand, the contingency planner generates a less conservative closed-loop trajectory than the nominal planner. We validate our planners using state-of-the-art trajectory prediction algorithms in autonomous driving simulators.

Keywords: Discrete event systems, Optimization, Manufacturing systems
Abstract: This paper presents a comparison between a continuous time mixed-integer linear programming (MILP) scheduling approach and a Monte-Carlo Tree Search (MCTS) based approach for scheduling a sequence-dependent, zero-buffer multi-batch process. The benchmark plant comprises shared inlet and outlet pipelines and parallel processing tanks, where tightly coupled transfers and conditional cleaning create strong resource interactions. For a consistent comparison, both methods are evaluated on the same benchmark with identical transfer and cleaning rules, including zero-buffer transfers and sequence-dependent cleaning. Performance is assessed over increasing demand levels in terms of makespan and computational behavior. For the tested small and medium demand scenarios, the MCTS approach attains the same makespan values as the MILP optimum, while the MILP computation time increases sharply with problem size and does not terminate within practical limits for the highest demand instance considered. In comparison, the MCTS approach returns feasible schedules within a fixed runtime budget and maintains near-optimal makespans as demand increases. The results provide valuable insights into the practical trade-off between solution quality guarantees and computational scalability and provide guidance for the selection of scheduling methods for industrial batch processes.

Keywords: Manufacturing systems, Optimization, Process control
Abstract: Advanced control methods have proven effective for controlling cavity pressure, a key determinant of part-quality attributes, in the plastics injection molding process. However, the abstract nature of the resulting control laws makes them difficult to interpret in a production environment, thereby limiting adoption in industrial applications. Additionally, controller optimization poses a severe challenge due to the diversity of mold geometries and materials. We propose a method to automatically optimize interpretable controllers during manufacturing while being cycle-efficient and risk-aware. The approach uses a Physics-Inspired Neural Mixture-of-Local-Experts model of the injection molding dynamics and augments its simulated closed-loop costs with a residual Gaussian Process, enabling Local Bayesian Optimization of controller parameters. We benchmark the algorithm against Vanilla Bayesian Optimization (BO) in simulation, using three controllers with parameter counts ranging from 1 to 30. Using the local method, we identify controller parameters that yield costs comparable to or lower than those of global BO over 20 optimization iterations, while mitigating high-cost excursions during tuning.

Keywords: Mining, minerals and petroleum, Predictive control, Optimization
Abstract: Mine dewatering is a safety-critical and energy-intensive process in underground mining, commonly implemented as a multilevel chain of reservoirs and pumping stations. Many installations still rely on level-threshold, rule-based control that maintains safe reservoir bounds but does not exploit flexible electricity prices. This paper investigates price-aware control for a multilevel mine dewatering use case using a physics-based MATLAB/Simulink model parameterized from industrial specifications.

We propose model predictive control (MPC) strategies designed to minimize electricity costs while ensuring reservoir safety constraints and adhering to discrete pump actuation limits. Our formulation models the nonlinear behavior of pumps and on-off decisions as a mixed-integer nonlinear program (MINLP). As a more computationally efficient alternative, we propose a linear approximation of the nonlinear flow mapping, resulting in an linear programming (LP) based MPC variant. We also model and assess the use of variable-frequency drives with the pumps. The results indicate that price-aware optimization can reduce dewatering energy cost almost 30%.

Keywords: Process control, Control applications, Cybersecurity
Abstract: Industrial control systems rely on alarm systems to alert operators to abnormal conditions. In practice, alarm systems frequently generate alarm floods - periods of rapidly occurring alarms that overwhelm operators and degrade situational awareness. Alarm floods may arise from conventional operational faults, such as equipment failures, or from deliberate cyberattacks designed to manipulate process behavior. Although both manifest as intense alarm activities, fault-induced and cyberattack-induced alarm floods are fundamentally different and require distinct diagnostic and response strategies.

This paper addresses the dual problem of online alarm flood classification and detection of previously unseen cyber intrusions. A unified deep learning framework based on a Class-Conditioned Autoencoder (C2AE) is proposed, which employs convolutional neural networks in both its encoder and decoder to process spatiotemporal alarm heatmaps. The framework jointly performs closed-set classification of known alarm flood categories and open-set recognition by identifying alarm patterns inconsistent with known operational faults. The proposed framework is validated on the Tennessee Eastman Process benchmark, where it achieves high accuracy in classifying operational faults and reliably detects cyberattacks as out-of-distribution alarm floods.

Keywords: Real-time systems, Optimization, Nonlinear systems
Abstract: Heavy-duty hydraulic manipulators (HHMs) operate under strict physical and safety-critical constraints due to their large size, high power, and complex nonlinear dynamics. Ensuring that both joint-level and end-effector trajectories remain compliant with actuator capabilities—such as force, velocity, and position limits—is essential for safe and reliable operation, yet remains largely underexplored in real-time control frameworks. This paper presents a nonlinear model predictive control (NMPC) framework designed to guarantee constraint satisfaction throughout the full nonlinear dynamics of HHMs, while running at a real-time control frequency of 1~kHz. The proposed method combines a multiple-shooting strategy with real-time sensor feedback, and is supported by a robust low-level controller based on virtual decomposition control (VDC) for precise joint tracking. Experimental validation on a full-scale hydraulic manipulator shows that the NMPC framework not only enforces actuator constraints at the joint level, but also ensures constraint-compliant motion in Cartesian space for the end-effector. These results demonstrate the method’s capability to deliver high-accuracy trajectory tracking while strictly respecting safety-critical limits, setting a new benchmark for real-time control in large-scale hydraulic systems.

Keywords: Cyberphysical systems, Mobile Robots, Neural networks
Abstract: Neural network controllers are increasingly deployed in robotic systems for tasks such as trajectory tracking and pose stabilization. However, their reliance on potentially untrusted training pipelines or supply chains introduces significant security vulnerabilities. This paper investigates backdoor (Trojan) attacks against neural controllers, using a differential-drive mobile robot platform as a case study. In particular, assuming the robot's tracking controller is implemented as a neural network, we design a lightweight, parallel Trojan network that can be embedded within it. This malicious module remains dormant during normal operation but, upon detecting a highly specific trigger condition defined by the robot's pose and goal parameters, compromises the primary controller's wheel velocity commands, resulting in undesired and potentially unsafe robot behaviours. We provide a proof-of-concept implementation of the proposed Trojan network, validated through simulation under two attack scenarios. The results confirm the effectiveness of the proposed attack and demonstrate that neural network-based robotic control systems are subject to potentially critical security threats.

Keywords: Fault-tolerant systems, Cyberphysical systems, Optimization
Abstract: We consider the problem of active fault-tolerant control in cyber-physical systems composed of strictly passive linear-time invariant dynamic subsystems. We cast the problem as a constrained optimization problem and propose an augmented primal-dual gradient dynamics-based fault-tolerant control framework that enforces network-level constraints and provides optimality guarantees for the post-fault steady-state operation. By suitably interconnecting the primal-dual algorithm with the cyber-physical dynamics, we provide sufficient conditions under which the resulting closed-loop system possesses a unique and exponentially stable equilibrium point that satisfies the Karush--Kuhn--Tucker (KKT) conditions of the constrained problem. The framework's effectiveness is illustrated through numerical experiments on a DC microgrid.

Keywords: Embedded systems, Process control, Real-time systems
Abstract: We implement an offset-free linear model predictive control on a programmable logic controller (PLC) following the IEC 61131-3 programming standard. We develop custom matrix operation routines to ensure efficient embedded implementation of a first-order primal-dual solver based on the inexact-Uzawa method. Using a real-life laboratory-scale quadruple tank experiment and a Beckhoff PLC, we demonstrate the effectiveness of the closed-loop system under TwinCAT runtime constraints.

Keywords: Embedded systems, Reinforcement learning, Cyberphysical systems
Abstract: Learning-enabled control systems increasingly rely on multi-agent reinforcement learning to operate in uncertain and interactive environments. While risk-aware decision-making has been shown to improve safety and robustness, deploying such algorithms on resource-constrained embedded platforms remains a significant challenge due to limited memory, compute, and communication resources. In this paper, we propose a risk-informed multi-agent reinforcement learning framework based on cumulative prospect theory (CPT) and introduce a CPT-SARSA algorithm suitable for real-time execution on low-power embedded hardware. We implement our algorithm on Arduino Nano 33 BLE Sense microcontrollers and evaluate real-time leader-follower coordination in a stochastic grid-world environment with obstacles and hazards. Our experiments demonstrate that CPT-informed agents achieve faster task completion and significantly fewer collisions compared to standard Q-learning, highlighting safety-efficiency tradeoffs under embedded constraints. This successful demonstration shows that complex multi-agent reinforcement learning algorithms can be deployed on low-cost commodity hardware, increasing access to state-of-the-art learning-enabled control without expensive compute resources.

Keywords: Cybersecurity, Cyberphysical systems, Estimation
Abstract: This paper presents an optimal secure estimation framework for linear Cyber-Physical Systems (CPS) based on an auxiliary system that provides redundant measurement information for the main plant. The output of the auxiliary system, designed with a specific dynamic structure, is combined with the plant output to generate a redundant information set transmitted over the communication channels. To increase the difficulty for attackers to identify the active communication channels, a Moving Target Defense (MTD) mechanism is implemented at the plant side prior to transmission. By introducing a switching mechanism among different modes of the output information that is a predefined set of observable-ensuring output combinations, the under-attack channel can be identified to further provide a secure estimation and resilient control on the Command and Control (C&C) side. The optimal design of the observers and controller is formulated by defining proper cost functions. The simulation results for the Vertical Take-Off and Landing (VTOL) aircraft are provided to demonstrate the effectiveness and performance of the proposed method.

Keywords: Predictive control, Identification, Kalman filtering
Abstract: This research presents a Nonlinear Model Predictive Control (NMPC) scheme integrated with a Line of Sight (LOS) guidance approach to address the path following problem of commercial inland vessels. The proposed method employs a quadratic running cost function to compute control actions that balance tracking performance and motion smoothness while adhering to the physical constraints imposed by the vessel’s maneuvering model. Moreover, a Recursive Least Squares (RLS) algorithm is incorporated to adapt the model parameters in real time based on the vessel's most recent operating conditions. The reference path is defined by waypoints forming straight line segments, which are subsequently used to determine the desired direction. The effectiveness of the proposed approach is validated through experiments conducted on a full-scale inland barge operating in a section of the Hollands Diep.

Keywords: Predictive control, Distributed parameter systems, Linear systems
Abstract: This paper proposes a Least Absolute Shrinkage and Selection Operator (LASSO)-based data-reduction framework for frequency-domain data-driven predictive control (FreePC) of infinite-dimensional linear time-invariant systems. By leveraging sparse modeling, we identify the essential frequency-domain information required to achieve target step responses, eliminating the need for complex first-principles modeling. The methodology is validated using an electromagnetic molding machine benchmark. Results indicate that the sparse FreePC framework is not only computationally efficient but also significantly more robust to measurement noise than conventional methods. Our findings suggest that identifying essential frequency-response characteristics through sparse modeling is a promising approach to the practical control of complex distributed-parameter systems without model reduction.

Keywords: Predictive control, Control applications, Energy Systems
Abstract: This paper introduces a Model-free Predictive Control (MFPC) scheme to optimize the hydrogen purity output in a nonlinear high-pressure alkaline electrolyzer. In contrast to conventional Model Predictive Control (MPC), the proposed MFPC approach eliminates the need for an explicit system model by employing the Model-free approach (an ultra-local model combined with real-time optimization). The controller adaptively compensates the electrolyzer for variations in electric current as well as pressurization and depressurization phases, while ensuring compliance with operational constraints. In four different operation scenarios, this study shows that MFPC exceeds standard Model-free Controllers (MFC), providing improved performance indicators such as reduced gas impurities, smoother control signals, and lower RMSE, ISE, IAE, and ITAE indices. Furthermore, similar to conventional MPC controllers, MFPC prevents constraint violations and also actuator overload, commonly encountered in MFC and MPC-based schemes, demonstrating its suitability for managing complex nonlinear processes.

Keywords: Predictive control, Learning, Optimization
Abstract: Mixed-integer model predictive control (MI-MPC) requires solving mixed-integer quadratic programs (MIQPs) online under strict timing constraints. Branch-and-bound is a standard solution method whose performance depends strongly on the branching rule. Unlike standard mixed-integer linear programming (MILP) benchmarks, MI-MPC induces a horizon structure where early-stage integer decisions couple to more of the predicted trajectory than late-stage decisions. This paper exploits this structure in two ways. First, it evaluates branching rules that prioritize early-stage variables. Second, it tests whether learning-to-branch methods developed for MILPs transfer to MI-MPC when trained to imitate a structure-aware expert. A novel horizon-aware hybrid branching rule, left-strong, is introduced and shown to reduce explored nodes, solve time, and worst-case latency relative to a state-of-the-art general-purpose baseline across two MI-MPC benchmarks. The results further show that strong branching, a common learning-to-branch expert, is ill-suited to this setting. Motivated by this finding, left-strong is used as an expert to train multilayer perceptron (MLP) and graph convolutional neural network (GCNN) policies by imitation learning. Both achieve high imitation accuracy, but only the MLP yields modest speedups due to lower inference overhead. Overall, the study shows that structure-aware branching captures most of the attainable branching gains in MI-MPC, leaving limited headroom for learning compared to MILP benchmarks with little structure.

Keywords: Predictive control, Linear robust control, Numerical analysis
Abstract: This paper derives an explicit and closed-form upper bound on the Hausdorff distance between the finite-horizon truncated minimal Robust Positively Invariant (mRPI) set and the infinite-horizon mRPI. Existing approaches rely on geometric convergence arguments or norm-based conditions and therefore do not provide a computable expression for the truncation error. We show that the Hausdorff gap between the truncated and infinite mRPI sets decays at a rate determined by the induced contraction factor of the system and that the bound depends only on this contraction rate and the disturbance magnitude. The resulting formula provides a numerically efficient and fully explicit estimate of the truncation error, enabling quantitative assessment of truncation accuracy in robust and tube-based MPC schemes. We further show that appropriate norm design can significantly accelerate the decay of the bound. Numerical experiments in moderate and high dimensions validate the correctness, tightness, and scalability of the proposed bound.

Keywords: Predictive control, Nonlinear systems, Energy Systems
Abstract: Model predictive control effectively manages constraints with robust performance, yet online implementation of nonlinear MPC remains a challenge for stiff, nonlinear and high dimensional systems. A prime example is a hybrid vehicle thermal management system (TMS) that integrates thermal energy storage (TES), thereby utilizing both conventional heat exchangers and phase change materials for cooling. As vehicle electrification imposes stricter performance demands on these systems, the need for advanced control strategies increases. This presentation introduces the design and real-time implementation of a 77-state MPC on an experimental hybrid TMS testbed [1]. We demonstrate that despite the system’s stiffness and scale, an explicit integration scheme can be formulated by extracting an appropriate linear system at each step of the online prediction horizon. A key advantage of this method is the minimal additional computational cost required to evaluate first-order gradients. Simulated and experimental results demonstrate the efficacy of this approach and the value of the TES in rejecting highly transient heat load disturbances.

[1] D. Gulewicz, U. Inyang-Udoh, T. Bird and N. Jain, "Nonlinear Model Predictive Control of a Hybrid Thermal Management System," in IEEE Transactions on Control Systems Technology, vol. 34, no. 2, pp. 892-905, March 2026, doi: 10.1109/TCST.2025.3646704.

Keywords: Cooperative control, Transportation systems, Simulation
Abstract: With the increasing deployment of autonomous transportation systems, longitudinal vehicle control has so far mostly been addressed using decentralized feedback-based approaches such as Adaptive Cruise Control (ACC). These methods offer low computational complexity and inherent robustness, and are widely deployed in production vehicles as part of Advanced Driver Assistance Systems (ADAS). However, in complex traffic situations, such as urban intersection scenarios, decentralized approaches often cannot fully exploit the available degrees of freedom. Fully autonomous traffic environments therefore motivate centralized control concepts that leverage system-level controllability and observability to coordinate vehicle behavior at intersections.

This paper presents a centralized longitudinal control approach designed for operation in both fully autonomous and mixed traffic at urban intersections and compares it to a decentralized reference in terms of travel speed and energy consumption. The centralized formulation addresses mixed traffic conditions by operating under partial controllability and limited predictability of non-autonomous vehicles. While fully autonomous scenarios allow the approach to exploit system-wide controllability, its behavior in mixed traffic is shaped by the specific traffic situation and the accuracy of vehicle motion predictions. The results illustrate how partial loss of controllability influences system-level performance.

Keywords: Switched systems, Automotive applications, Autonomous systems
Abstract: Hybrid electric vehicles (HEVs) often require transitions between multiple energy sources, such as the internal combustion engine (ICE) and electric motor (EM). These abrupt transitions, known as zero dwell-time switching, can cause drivability problems, torque discontinuities, and reduced efficiency. This paper proposes an adaptive dwell-time switching strategy to alleviate these issues. The HEV is modeled as a switched system, and stability analysis is conducted for each mode and the dwelling period. We present a comparison of zero, fixed, and adaptive dwell-time switching in the HEV powertrain over Federal Test Procedure (FTP-75) drive cycle.

Keywords: Transportation systems, Control applications, Robotics applications
Abstract: This paper presents simulation and experimental validation of a backstepping-based time-gap control strategy for autonomous vehicle platooning using a small-scale multi-robot platform. The controller, originally developed in [1] for nonlinear vehicle platoon dynamics, is implemented on a three-robot e-puck2 platoon consisting of one leader and two followers. A Python simulation is first used to reproduce the theoretical controller behavior under idealized conditions, followed by a Webots physics-based simulation that incorporates realistic effects such as discrete sampling, sensor noise, and actuator limitations. Both simulations demonstrate accurate regulation of speed and time-gap errors and confirm the stability properties predicted by the theoretical controller. Physical experiments conducted using e-puck2 robots on a guided track further demonstrate that stable platoon formation can be maintained under realistic sensing, communication, and actuation constraints, although larger tracking errors are observed due to sensor inaccuracies, communication latency, and motor response limitations. These results demonstrate the feasibility of using small-scale robotic platforms for validating nonlinear platooning controllers and provide a reproducible pathway from theoretical design to physical implementation.

Keywords: Transportation systems, Cybersecurity
Abstract: Autonomous vehicle platoons depend on directed communication for coordination, but are highly susceptible to Denial-of-Service (DoS) attacks that undermine stability and cooperation. This work addresses the challenging case in which both the leader and a follower are simultaneously subjected to DoS attacks. A resilient vehicular control strategy is proposed to detect, mitigate, and recover from such adversarial conditions. Attack detection is achieved through a divergence-based scheme that employs dual incremental timers to estimate the time-varying delay induced by the attack, using reference trajectories. Once the delay is quantified, a resilient controller is activated to restore consensus, regulate attacked vehicle dynamics, and ensure safety constraints. To preserve network functionality during prolonged disruptions, a switching mechanism reassigns the compromised leader as a follower and designates a new leader, thereby maintaining hierarchical structure and continuous guidance. The new leader is selected using a rule-based strategy that evaluates communication connectivity, spatial proximity, and motion similarity over short-horizon measurements relative to the former leader. Distributed control preserves consensus in the platoon under attack conditions. Simulation studies validate the proposed strategy, demonstrating its robustness and practical significance for secure platooning in intelligent transportation systems.

Keywords: Transportation systems, Game theory, Optimization
Abstract: In this paper, we propose an integrated matching and pricing algorithm for multi-passenger ride-sharing systems. While ride-sharing is a promising solution for reducing traffic congestion and CO2 emissions, existing mechanisms, particularly the Vickrey-Clarke-Groves (VCG) mechanism, often fail to ensure budget feasibility (i.e., they may result in a budget deficit), which poses a significant challenge to sustainable operation. To address this issue, we formulate the matching and pricing problem as a single optimization problem incorporating a novel budget-feasibility constraint that explicitly prevents system deficits. We prove that the proposed mechanism simultaneously satisfies three critical economic properties: strategy-proofness, individual rationality, and budget feasibility. Furthermore, through numerical simulations conducted on a real-world road network in Yokohama, Japan, demonstrate that the proposed method consistently guarantees a budget surplus even in scenarios where the standard VCG mechanism incurs a deficit, with only a marginal loss in overall social efficiency.

Keywords: Automotive applications, Nonlinear systems
Abstract: This paper presents the design and analysis of a safe lane-keeping controller based on a kinematic single-track vehicle model. A control barrier function (CBF) is derived for the lane-keeping problem using geometrical considerations. This is used to design a safety filter that can be applied on top of a nominal lane keeping controller and guarantees that the vehicle remains within the lane boundaries. The effects of applying the safety filter to the controlled vehicle are thoroughly analyzed using a series of numerical simulations and phase portraits, including changes in the global dynamics, passenger comfort and input signals. The results are validated with experiments on a test vehicle that verify that the safety filter is able to keep the vehicle inside the lane boundaries. The results highlight the benefits of safety filters in lateral vehicle control, and can be used as a basis for the design of more complex controllers involving higher fidelity vehicle models to enhance the safety of both human-driven and automated vehicles.

Keywords: Estimation, Observers, Robotics
Abstract: This paper studies nonlinear observer design for rigid-body extended pose estimation using inertial measurements and generic exteroceptive sensing. The estimation problem is formulated as a cascade architecture that separates translational dynamics from rotational kinematics while preserving the geometric constraint of attitude evolution on SO(3). By embedding the inertial navigation model into a Linear Time-Varying (LTV) representation, we construct an observer composed of a Kalman-Bucy-type estimator for translational states and an auxiliary unconstrained attitude variable, coupled with a nonlinear geometric reconstruction filter evolving on SO(3). The cascade interconnection is analyzed within a nonlinear systems framework where uniform observability of the LTV subsystem guarantees almost global asymptotic stability of the overall nonlinear observer. For a benchmark GPS–landmark-aided configuration, explicit sufficient conditions on admissible trajectories are derived to ensure uniform observability. Simulation results illustrate the effectiveness of the proposed estimation framework.

Keywords: Estimation, Energy Storage, Automotive applications
Abstract: Accurate battery module State-of-Health (SoH) estimation remains challenging due to coupled electrical–thermal dynamics and the influence of active thermal management. While thermal diagnostic methods such as Differential Thermal Voltammetry (DTV) and Differential Heat Flux Voltammetry (DHFV) have demonstrated strong performance at the cell level and initial applicability to modules, they are used on laboratory testbenches and rely on the thermal response of isolated components. In real-world settings where battery thermal management systems (BTMS) are active, the batteries are not thermally isolated, therefore these thermal diagnostic tools are no longer applicable. This paper proposes Thermal Energy Voltammetry (TEV), a novel thermal-based indirect analysis method that integrates temperature and heat flux measurements through a heat balance at the system level to reconstruct generated thermal energy, exploiting its invariance to cooling actuation. Using a multi-fidelity virtual testbed incorporating electrochemical–thermal dynamics, fluid flow, electrical interconnections, and feedback control, TEV is evaluated at both the cell and module levels under open-loop and closed-loop thermal management. Results demonstrate that TEV preserves consistent health-indicating features across varying cooling regimes for cells and homogeneous modules with parallel-connected cells. The effects of cell-to-cell (CtC) variation are also explored and their impacts on SoH estimation are investigated.

Keywords: Estimation, Control applications, Sensors
Abstract: Fiber-optic sensing systems enable applications such as pipeline monitoring where accurate state estimation from optical measurements is critical. Intensity-domain interference measurements are widely used in coherent optical sensing systems, yet polarization-dependent effects are often neglected or treated as unmodeled disturbances, introducing structured modeling errors and limiting estimation accuracy. This paper presents a nonlinear model that explicitly incorporates polarization-dependent interference using intensity measurements from three polarization projections. A seven-parameter optical state representation is introduced to parameterize polarization-component amplitudes, relative pulse phase, and polarization-dependent phase differences. Based on this grey-box model, where the structure is derived from physics and parameters are estimated from data, a nonlinear least-squares framework is developed to recover the optical state using measurements from a tri-state polarization diversity receiver (PDR) with the three polarization projection angles. Experimental results demonstrate accurate reconstruction of measured intensity projections across all projection angles. Acoustic single-tone validation confirms that estimated state parameters track physically meaningful perturbations: the applied tone frequency is recovered via Fast Fourier Transform (FFT) of the estimated relative phase parameter. This work bridges polarization-aware optical modeling and practical intensity-domain measurements by enabling polarization state estimation using intensity-only data.

Keywords: Sensor networks, Mobile Robots, Estimation
Abstract: Accurate forecasting of future solar irradiance is essential for the effective control of solar thermal power plants. Although various kriging-based methods have been proposed to address the prediction problem, these methods typically do not provide an appropriate sampling strategy to dynamically position mobile sensors for optimizing prediction accuracy in real time, which is critical for achieving accurate forecasts with a minimal number of sensors. This paper introduces a dissimilarity map derived from a kriging model and proposes a persistent coverage control algorithm that effectively guides agents toward regions where additional observations are required to improve prediction performance. By means of experiments using mobile robots, the proposed approach was shown to obtain more accurate predictions than the considered baselines under various emulated irradiance fields.

Keywords: Ships and offshore vessels, Identification, Mechatronic systems
Abstract: Reliable motion models are essential for advanced automation functions of inland waterway vessels, particularly for low-speed maneuvers in confined environments such as lock approaches. This work proposes a systematic, maneuver-based framework for full-scale parameter identification of inland vessels. The system dynamics are described by a nonlinear maneuvering model comprising surge, sway, and yaw dynamics.

A structured identification procedure is employed in which the vessel dynamics are excited sequentially using dedicated maneuvers designed to isolate specific motions, enabling accurate identification of both hydrodynamic and actuator thrust coefficients. A nonlinear optimization problem is formulated, identifying the parameters using maneuver data generated by a reference model. The simulation study demonstrates high prediction accuracy in all identification maneuvers and validates the identified parameters on independent maneuvering trajectories involving combined excitation of the complete vessel dynamics.

Keywords: Identification, Estimation, Numerical analysis
Abstract: Maximum likelihood estimation for high-dimensional linear Gaussian state-space models via the Expectation-Maximization algorithm becomes computationally prohibitive, requiring cubic complexity in the state dimension and quadratic memory per iteration. We propose a scalable approximation framework that exploits banded structure in transition matrices through covariance truncation in the expectation step and matrix-free Krylov subspace methods in the maximization step. The approach is analyzed within the inexact EM framework, establishing convergence to stationary points with controlled approximation error. Experiments with state dimensions reaching two thousand demonstrate feasibility where standard methods fail, achieving near-oracle prediction accuracy with memory reductions exceeding two orders of magnitude.

Keywords: Cybersecurity, Cyberphysical systems, Navigation
Abstract: This paper investigates resilient positioning, navigation, and timing for safety‑critical maritime navigation systems subject to false‑data injection cyber-attacks on inertial measurement units and magnetometers, in the presence of potentially degraded global navigation satellite system measurements. We propose a resilient, loosely coupled error‑state navigation filter specifically tailored to surface vessels' motion constraints. The filter integrates (i) an unknown‑input‑decoupled error‑state update designed to mitigate IMU data injections and (ii) an adaptive fading‑memory, two‑stage separation architecture capable of tracking nonstationary integrity degradation in magnetometer and GNSS measurements. Simulation studies employing a representative surface-vessel dynamics model and incorporating cyber-attack scenarios targeting IMU, magnetometer, and GNSS position measurements demonstrate the resilience of the proposed navigation filtering framework for marine vessels.

Keywords: Autonomous systems, Navigation, Predictive control
Abstract: Efficient autonomous navigation in a priori unknown, cluttered environments remains challenging due to limited sensing, unknown obstacles, and the need to respect dynamic and safety constraints during motion. Existing exploration strategies often decouple environment discovery from control, resulting in suboptimal coverage, excessive replanning, or unsafe trajectories. This work addresses this gap by introducing a framework for autonomous exploration and navigation in a priori unknown, cluttered environments using Model Predictive Control (MPC) coupled with Heat Equation Driven Area Coverage (HEDAC) biased Rapidly-exploring Random Trees (RRT). This method incrementally maps unknown spaces while safely avoiding obstacles through convex corridor constraints, and transitions from exploration to goal-directed navigation. Numerical simulations in a cluttered environment demonstrate that the HEDAC bias achieves real-time, adaptive, and collision-free navigation while improving coverage efficiency and reducing redundant path generation compared to a frontier-biased sampling strategy. These results provide insights into how exploration biasing impacts closed-loop control performance and execution safety in unknown environments.

Keywords: Robotics applications, Nonlinear systems, Navigation
Abstract: Ensuring the safety of dynamical systems is crucial, where collision avoidance is a primary concern. Recently, control barrier functions (CBFs) have emerged as an effective method to integrate safety constraints into control synthesis through optimization techniques. However, challenges persist when dealing with convex primitives and tasks requiring torque control, as well as the occurrence of unintended equilibria. This work addresses these challenges by introducing a high-order CBF (HOCBF) framework for collision avoidance among convex primitives. We transform nonconvex safety constraints into linear constraints by differentiable optimization and prove the high-order continuous differentiability. Then, we employ HOCBFs to accommodate torque control, enabling tasks involving forces or high dynamics. In addition, we analyze the issue of spurious equilibria in highorder cases and propose a circulation mechanism to prevent the undesired equilibria on the boundary of the safe set. Finally, we validate our framework with three experiments on the Franka Research 3 robotic manipulator, demonstrating successful collision avoidance and the efficacy of the circulation mechanism.

Keywords: Localization, Navigation, Marine/underwater robotics
Abstract: Real-Time Kinematic (RTK) augmentation and Inertial Navigation System (INS) integration are widely used to improve Global Navigation Satellite System (GNSS) positioning performance. However, on inland waterways, bridges and surrounding structures can degrade satellite visibility and correction availability, causing RTK augmentation loss, and GNSS/INS fusion transients. Since these effects depend on the local environment and sensor configuration, nominal receiver specifications are insufficient, and deployment-specific characterization is required.

This paper presents a benchmarking study of an AsteRx-i3 D Pro+ GNSS/INS receiver installed within the mobile Sensor Box developed at KU Leuven. The study combines a real-world bridge-passage case study, static benchmarking, and closed-loop path-following experiments. The static benchmarking evaluates four receiver configurations: standalone GNSS, standalone GNSS with INS integration, RTK-augmented GNSS, and RTK-augmented GNSS with INS integration. The closed-loop experiments use INS-integrated GNSS as the navigation input and compare path-following operational performance with and without RTK augmentation.

Results show that correction loss during bridge passage causes reduced positioning accuracy, increased positioning uncertainty and recovery-induced state jumps exceeding 1m. Static benchmarking and closed-loop experiments confirm that RTK augmentation substantially improves positioning precision and uncertainty consistency, while INS integration supports short-term continuity during RTK unavailability but may introduce drift, bias, or transient uncertainty variations. By characterizing the deployment-specific receiver behavior with RTK augmentation and INS integration, this study motivates higher-level state estimation as a necessary next step toward spatially continuous and uncertainty-consistent positioning on inland waterway. The experimental data are released at: https://ssrn.com/abstract=6372458 to support reproducibility and further research.

Keywords: Nonlinear systems, Navigation, Aerospace applications
Abstract: This paper investigates the problem of attitude and air velocity estimation for fixed-wing unmanned aerial vehicles (UAVs) using IMU measurements and at least one Pitot tube measurement, with almost global asymptotic stability (AGAS) guarantees. A cascade observer architecture is developed, in which a Riccati/Kalman-type filter estimates the body-fixed frame air velocity and the vehicle’s tilt using IMU data as inputs and Pitot measurements as outputs. Under mild excitation conditions, the resulting air velocity and tilt estimation error dynamics are shown to be uniformly observable. The estimated tilt is then combined with magnetometer measurements in a nonlinear observer on SO(3) to recover the full attitude. Rigorous analysis establishes AGAS of the overall cascade structure under the uniform observability (UO) condition. The effectiveness of the proposed approach is demonstrated through validation on real flight data.

Keywords: Mechatronic systems, Nonlinear systems, PID control
Abstract: This study introduces a modified version of the Constant-in-Gain, Lead-in-Phase (CgLp) filter, which incorporates a feedthrough term in the First-Order Reset Element (FORE) to reduce the undesirable nonlinearities and achieve an almost constant gain across all frequencies. A backward calculation approach is proposed to derive the additional parameter introduced by the feedthrough term, enabling designers to easily tune the filter to generate the required phase. The paper also presents an add-on filter structure that can enhance the performance of an existing LTI controller without altering its robustness margins. A sensitivity improvement indicator is proposed to guide the tuning process, enabling designers to visualize the improvements in closed-loop performance. The proposed methodology is demonstrated through a case study of an industrial wire bonder machine, showcasing its effectiveness in addressing low-frequency vibrations and improving overall control performance.

Keywords: Computational methods, Optimization, Complex systems
Abstract: We study a constraint-preserving Quantum Approximate Optimization Algorithm (QAOA) for the multi-agent task allocation problem under injective allocation constraints. The allocation cost is encoded into a diagonal cost Hamiltonian derived from the given cost matrix, while feasibility is enforced by a mixer Hamiltonian that restricts transitions along the edges of the feasible-state graph, thereby preventing probability leakage to infeasible allocations. We further provide a graphtheoretic sensitivity bound showing that the mixer unitary’s β- perturbation is bounded by the maximum degree of the feasiblestate graph, clarifying how graph connectivity impacts mixing dynamics. Numerical simulations on a toy instance using Qiskit validate that the proposed approach concentrates sampling probability on low-cost feasible allocations.

Keywords: Reinforcement learning, Control Technology, Verification and validation
Abstract: HVAC systems represent a major share of building energy consumption. Traditional control strategies are limited in coordinating energy-comfort tradeoffs across multiple zones simultaneously. Reinforcement learning (RL) offers adaptive, data-driven control that optimizes performance over time. How- ever, deploying learned neural network controllers in safety- critical building systems remains challenging due to lack of formal safety guarantees. We propose a safety-certified deep RL framework for multi-zone residential HVAC control. Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) agents are trained in an EnergyPlus/Sinergym simulation to minimize energy consumption while maintaining thermal comfort. Post- training safety certification is performed on the PPO policy using Lipschitz-based forward invariance analysis, building on existing tools for the computation of Lipschitz constants for neural networks, to guarantee constraint satisfaction. Both agents are evaluated over an annual simulation cycle in an eight-zone variable refrigerant flow (VRF) testbed. The PPO agent achieves 67% comfort violation reduction compared to rule-based control, while the SAC agent achieves 27.6% energy savings. The PPO policy satisfies formal safety certification with a margin of 2.003◦C. These results demonstrate the feasibility of combining reinforcement learning with post-training safety verification for multi-zone building control.

Keywords: Reinforcement learning, Machine learning, Robotics
Abstract: We present a concise summary of disturbance-aware reinforcement learning (DiAReL), a method for improving zero-shot sim2real policy transfer in robot control by explicitly estimating and conditioning on additive disturbances in delayed environments [1].

[1] M. Malmir, J. Josifovski, N. Klarmann, and A. Knoll, “Diarel: Reinforcement learning with disturbance awareness for robust sim2real policy transfer in robot control,” IEEE Transactions on Control Systems Technology, vol. 34, no. 2, pp. 1037–1043, 2026.

Keywords: Reinforcement learning, Control applications, Markov processes
Abstract: Industrial refrigeration facilities are among the most energy-intensive industrial systems and operate under strict temperature constraints to ensure product safety and quality. Optimizing energy consumption can therefore lead to substantial cost savings. The allowable temperature flexibility of stored products effectively provides thermal storage, which can be leveraged for intelligent control. However, in facilities with multiple rooms coupled through a centralized cooling unit, commonly employed heuristic strategies are often insufficient for meaningful cost reduction. More comprehensive model-based control approaches, e.g., model predictive control, also suffer from uncertainty in thermal parameters, external disturbances, and product characteristics. This paper explores the potential of model-free control in industrial refrigeration. We propose a reinforcement learning–based control strategy that learns an energy-efficient cooling policy through interaction with the system, without requiring explicit knowledge of the dynamics. Simulation results on a simplified facility demonstrate that the proposed approach exploits both thermal storage and inter-room coupling to reduce power consumption while maintaining temperature constraints and outperforming heuristic controllers. These positive results suggest that further study is warranted.

Keywords: Health and medicine, Fuzzy control, Nonlinear robust control
Abstract: Effective blood glucose regulation in individuals with Type I diabetes remains a challenging control problem due to strong nonlinearities, inter-patient variability, and unmodeled disturbances such as physical activity. Exercise introduces rapid and highly uncertain changes in insulin sensitivity, often leading to hypoglycemic events when inadequately handled.'lb address these challengcs, this paper pr(,poscs :rn rxcrrisc-au'arc artifir:ial pancreas framework based on a R.einforcement I-earning (RL) tuned Hybrid Type-3 fuzzy logic controller. The proposed con- troller explicitly models higher-order uncertainty through Type-3 fuzzy membership functions, enabling improved representation of intra and inter-day physiological variability compared to Type- 1 and Type-2 fuzzy systems. An established exercise model is adopted to design and test the controller. Monte Carlo simulations are performed to demonstrate robustness under wide-ranging metabolic uncertainties and physical activity conditions. Simu- Iation results demonstrate that the proposed Rl-tuned Hybrid Type-3 fuzzy controller effectively converges glucose levels to the desired value and sigllificantlv irnpro,r'es gluurse regulation performance, reducing both hyperglycemic and hypoglycemic events while maintaining stable insulin delivery during post meal and exercise periods exhibiting enhanced robustness, adaptability, and safety. These results suggest that combining exercise model- ing with reinforcement learning optimized Type-3 frtzzy control offers a promising pathway toward reliable and patient-resilient artificial pancreas system.

Keywords: Aerial robotics, Cooperative control, Optimization
Abstract: This study addresses the challenge of guaranteeing safety for multi-UAV cooperative transport systems operating under disturbances such as wind. Because existing robust Control Barrier Functions (CBF) like Input-to-State Safety CBF (ISSf-CBF) are typically limited to first-order systems, applying them to second-order UAV dynamics causes instability. To overcome this, we utilize an ISSf High-Order Control Barrier Function (ISSf-HOCBF) generalized for systems with a relative degree of two. The primary contribution of this work is the development and experimental implementation of a unified framework that explicitly integrates ISSf-HOCBF to achieve disturbance-robust obstacle avoidance, while employing standard HOCBF for maintaining inter-agent formation.

Keywords: Aerospace applications, Navigation, Aerial robotics
Abstract: Low-altitude airspace operations are expanding rapidly, intensifying safety requirements for uncrewed aerial vehicles (UAVs) navigating cluttered and time-varying environments. This paper advances the artificial potential field (APF) method for collision avoidance through three variants that enhance geometric feasibility, incorporate spatiotemporal risk, and escape local minima. The proposed framework integrates a baseline with goal-aligned anisotropy, a temporal variant with sector-wise Time-to-Collision (TTC) weighting and safety-barrier potentials for preemptive responses, and a topological variant that leverages a histogram-based gap selector to break symmetric force equilibria. Together, these designs reconcile sensing constraints, early evasion, and symmetry breaking. To validate robustness, the proposed approach is evaluated through Monte Carlo simulations within randomized three-dimensional (3D) scenarios populated with mixed threats. Quantitative results confirm that the proposed gap-guided method avoids stagnation in the tested scenarios, achieving a 100% success rate across all trials, thereby demonstrating reliable avoidance of static obstacles, fast-moving intruders, and deadlock traps.

Keywords: Aerospace applications, Reduced order modeling, Predictive control
Abstract: Accurate and safe thrust control of rocket engines requires regulating the combustion chamber pressure and oxidizer-to-fuel ratio within strict operational limits. This paper presents a convex successive linear Model Predictive Control (SLMPC) framework for setpoint tracking of these variables in a pressure-fed rocket engine under actuator and state constraints. A lumped parameter dynamic model is derived and subsequently reduced by neglecting fast fluid dynamics. This yields a two-state representation that preserves the dominant engine dynamics while offering reduced computational complexity when employed for MPC. The resulting SLMPC formulation achieves accurate pressure and mixture-ratio control. Simulation results demonstrate that the proposed approach outperforms a conventional PI controller in tracking speed and constraint satisfaction, while closely matching the performance of a nonlinear MPC. The SLMPC requires 6 ms per solve at 30 Hz, which is six times faster than the nonlinear MPC.

Keywords: Estimation, Aerospace applications
Abstract: Use of quadrotor UAVs for wind velocity estimation is gaining popularity in recent studies, leveraging their maneuverability, compact size and low cost. Among available approaches, model-based wind velocity estimation is most commonly used, since it relies only on onboard sensors. However, as the quadrotor is a nonlinear system, this task becomes challenging. This study evaluates a discrete quadrotor dynamic model, formulated on SE(3) using a Lie Group Variational Integrator, and compares its wind-estimation performance against the commonly used continuous model. The discrete model avoids the integration residuals present in continuous formulations. By operating directly on SE(3), both the continuous and discrete formulations avoid singularities associated with Euler angles and ambiguity present with quaternions. Wind velocity estimation is performed with both models applied to an Extended Kalman filter (EKF) and an Unscented Kalman filter (UKF). Each algorithm is evaluated in simulations and real-time outdoor flights. The numerical results indicate that the discrete dynamics coupled with UKF achieve higher estimation accuracy than EKF, when used with cost-effective sensors. Real-time flight data analysis is consistent with the numerical simulation results, showing that both continuous-time and discrete-time UKF exhibit higher estimation accuracy. These findings demonstrate the effectiveness of discrete-time quadrotor models with UKF, while highlighting the limitations of EKF in the presence of a mismatch between the model and the quadrotor.

Keywords: Fault detection/accomodation, Aerial robotics, Estimation
Abstract: Propeller defects are among the most common safety-critical failures in multirotor Unmanned Aerial Vehicles (UAVs). Since such defects alter the control effort required to maintain flight, this paper presents a computationally lightweight fault detection and isolation method using standard onboard signals, including Electronic Speed Controller (ESC) duty cycle, battery voltage, and controller thrust and torque commands. A nominal thrust-voltage map is first identified from healthy trajectories. Static per-motor efficiency parameters are then estimated for each trajectory using a regularized least-squares formulation. The resulting relative efficiency vector is treated as compositional data: a centered log-ratio transform maps it to log-ratio coordinates, where a Mahalanobis-distance test is used for fault detection. For trajectories flagged as faulty, rotor-level isolation is performed using per-motor (w)-scores relative to healthy statistics. On a limited benchmark dataset with five healthy and nine faulty trajectories spanning edge cuts, cracks, and surface-unbalance faults, all faults were detected with no observed false alarms, and the isolation results predominantly identified the correct faulty motor.

Keywords: Aerial robotics, Sliding mode control, Robotics applications
Abstract: A coaxial tilt-rotor Unmanned Aerial Vehicle (CTR-UAV) featuring two front tilting coaxial rotors and a rear fixed rotor can hover at arbitrary pitch angles and generate thrust in arbitrary directions through coordinated pitch and yaw. This capability makes it well suited for fixed-attitude aerial physical interaction tasks.This paper proposes a robust control scheme for a CTR-UAV to achieve constant force regulation for dedicated aerial pushing tasks. To address the unknown interaction forces and wind disturbances,a high-order sliding-mode observer (HOSMO) is employed to estimate lumped disturbance force and torque in finite time. Based on estimation results, a super-twisting sliding-mode controller (STSMC) with an anti-windup mechanism is designed to maintain system stability and tracking performance when the tilt-angle limits are reached. Hardware-in-the-loop (HIL) experiments demonstrate that the proposed scheme achieves contact force regulation with a fixed-attitude in the presence of wind disturbances and actuator constraints.

Keywords: Communication networks, Control Technology, Game theory
Abstract: Dynamic bandwidth sharing in Satellite Communication (SATCOM) is challenging due to limited link capacity, finite terminal buffer sizes, and delayed bandwidth requests. Under high-load conditions, these factors can jointly lead to queue buildup and traffic loss. Many SATCOM systems rely on centralized bandwidth management in which Remote Terminals (RTs) submit bandwidth requests that are processed by a Satellite System Controller (SSC) on an epoch basis. In practice, the request–allocation loop experiences non-negligible round-trip delay due to propagation, processing, transmission, and queueing, which can cause demand mismatch and degraded queue behavior. This paper studies a delay-aware SATCOM terminal-management framework in which each terminal uses a Linear Quadratic Regulator (LQR) to generate bandwidth requests, while the SSC allocates limited aggregate capacity across terminals. We implement a control-aware Shapley allocation method that formulates resource (bandwidth) allocation as a cooperative game and assigns rates according to each terminal’s marginal contribution, subject to capacity and request constraints. The approach is evaluated under heterogeneous per-terminal delay settings and compared with a baseline weight-based proportional allocation scheme. Using dropped traffic, queueing delay, utilization, and computational complexity as evaluation metrics, the study characterizes trade-offs between the two allocation strategies and identifies operating conditions under which each method is preferable.

Keywords: Communication networks, Cybersecurity, Cyberphysical systems
Abstract: Satellite communication networks (SATCOM) rely on periodic terminal-reported telemetry for resource allocation and supervisory control, making the system vulnerable to jamming that degrades the quality of the reported link state. This paper presents a reproducible telemetry generation and detection workflow for studying jamming detection. A MATLAB-based generator using the 5G New Radio (NR) Toolbox produces standards-consistent link telemetry for four terminals with 20,000 samples per terminal. Here, New Radio refers to the 3GPP 5G air interface, and the toolbox provides standards-aligned physical-layer procedures, such as Orthogonal Frequency Division Multiplexing (OFDM) numerology, channel models, and link adaptation that map channel quality to modulation and coding. The generated telemetry includes Signal-to-Interference-plus-Noise Ratio (SINR), throughput, Transport Block Size (TBS), Block Error Rate (BLER), and energy per bit. The link simulation serves as a telemetry emulator for terminal reporting, while temporal correlation is introduced via a drifting AR-type SNR process to model slowly varying channel conditions common in operational links. A bursty jamming scenario is injected into a single terminal by reducing the effective SNR using a jammer-to-noise ratio process, along with additional Additive White Gaussian Noise (AWGN)-like uncertainty to emulate variability in the receiver noise floor. The dataset also stores the modulation order and target code rate derived from a Channel Quality Indicator (CQI) to Modulation and Coding Scheme (MCS) mapping to capture the effect of link adaptation on reported performance. Using this telemetry, a Python implementation trains a Temporal Convolutional Network (TCN) window classifier that operates on sliding windows of reported Key Performance Indicators (KPIs) and modulation and coding indicators. The model uses dilated causal convolutions with residual gated blocks and attention-based temporal pooling to learn temporal signatures of jamming bursts. The resulting framework supports control-oriented analysis and provides a baseline for comparing learned telemetry monitors with simple threshold-based monitoring in the SATCOM terminal management settings.

Keywords: Communication networks, Kalman filtering, Sensor networks
Abstract: This work studies uplink radio resource allocation for wireless state estimation in a networked control loop, where multiple sensors share an IEEE 802.11ax UL-OFDMA channel. Our communication model accounts for discrete resource unit allocation, a finite set of MCS and transmit power choices, and packet success probabilities obtained from an EESM-based PHY abstraction. In this setup, the transmission time given by the PPDU duration is dictated by the slowest scheduled user. On the control side, we derive a control-aware value of information (VOI) metric from the LQR-induced Lyapunov drift. This VOI quantifies the weighted reduction of the Kalman error covariance that results from a successful packet delivery. Combining these ingredients, we formulate a per-cycle utility function that trades off the expected VOI against the added latency and transmit power costs, while satisfying UL-OFDMA feasibility constraints. We then propose a greedy scheduler enhanced with a no-starvation rule. To keep the computational overhead manageable, the scheduler is implemented with a lazy branch-and-bound scheme that reduces the number of per-cycle utility evaluations without altering the selection objective. Simulations on an unstable flexible beam benchmark under Rayleigh fading demonstrate that our Lyapunov priority scheduler achieves a lower average LQG stage cost than a Round Robin baseline, even when both policies use the same RU granularity, power budget, and channel realizations. These results underscore the advantage of coupling physical-layer decisions with control-driven estimation priorities.

Keywords: Sensors, Machine learning, Vision
Abstract: Accurate estimation of hopper dumping level is essential for effective supervisory and regulatory control in oil sands ore preparation plants. Specifically, the estimated level can be employed within a regulatory control mechanism to ensure safe hopper operation by preventing hopper overfilling events and apron feeder dry-run scenarios, while a supervisory control framework could be incorporated to optimize the dumping time. Vision-based sensing for hopper level estimation provides a non-intrusive alternative to conventional instrumentation; however, its reliability degrades under industrial conditions where oil-sands stain on the hopper walls and reduce visual separability between background hopper structure and foreground oil-sands material. Under such appearance-challenging conditions, classical image segmentation pipelines require extensive post-processing and remain highly sensitive to visual variability, resulting in unreliable interface extraction and inaccurate level estimates. This work investigates image-to-image domain translation as a means of improving interface extraction for vision-based hopper level soft sensing. We propose a task-oriented, boundary-aligned translation framework that directly maps raw RGB hopper images to a semantic material interface representation, which requires no post-processing and is explicitly optimized for level estimation and control. Experimental results on industrial hopper imagery demonstrate that the proposed approach produces interpretable interface representations and yields accurate level estimates suitable for closed-loop control, providing an efficient alternative to classical multi-stage vision pipelines.

Keywords: Kalman filtering, Nonlinear systems, Identification
Abstract: This article presents the design and implementation of an invariant extended Kalman filter (EKF) for wind estimation using a small, fixed-wing uncrewed aerial vehicle (UAV). First, it is shown that the UAV’s dynamics, extended to include a trivial model for the wind velocity dynamics, are invariant with respect to the Lie group SE(3), the special Euclidean group of rigid transformations. It is also shown that an output comprising the vehicle’s inertial state and airspeed is equivariant with respect to SE(3). Based on these analysis results, an invariant EKF is then designed for the extended state equations and is implemented on experimental flight data. Wind estimates from the invariant EKF are compared with estimates obtained using a conventional EKF for two nominal aircraft motions: constant altitude, wings level flight, and constant altitude turning flight. The obtained wind estimates are then compared with wind measurements reconstructed using data from the aircraft’s air data unit comprising an angle of attack (AoA) flow vane, flank angle flow vane, and Kiel probe. The experimental results corroborate the expectation that the invariant EKF outperforms a conventional EKF in wind estimation accuracy.

Keywords: Cooperative control, Sensor networks, Aerial robotics
Abstract: This paper is concerned with persistent coverage control of unmanned aerial vehicles (UAVs) based on an information decay mechanism. Persistent coverage control is a framework to drive a limited number of UAVs to patrol a specified environment based on the time-varying importance of each location. In the previous studies on the information-decay-based method, the importance map is determined solely by the past and current configuration of the UAVs. However, prior information on the environment, such as locations with a high number of casualties in a disaster area, is crucial in practical applications. In this paper, we propose a new persistent coverage control method by introducing a modified importance map that combines the prior environmental information and the information decay mechanism. The effectiveness of the proposed method is verified through numerical simulations with three types of information reliability functions.

Keywords: Distributed control, Cooperative control, Automotive applications
Abstract: Cooperative Adaptive Cruise Control (CACC) leverages Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communication in the control synthesis, enhancing platoon safety and coordination. This work develops an error-dynamics-based control framework that incorporates nonlinear aerodynamic drag and linear rolling resistance for both centralized and decentralized CACC. The centralized controller is augmented with networked predictive control to compensate communication delays and packet loss by predicting future states and transmitting time-stamped control sequences. The decentralized controller serves as a backup mode under degraded communication, relying on delayed predecessor data without delay compensation. A delay-dependent switching strategy transitions between centralized, blended, and decentralized modes during communication failures and employs estimator-state sharing to prevent drift during mode transitions. Simulation results demonstrate that centralized with delay compensation minimizes spacing and velocity errors compared to decentralized, while the hybrid framework ensures reliable operation under adverse delay conditions.

Keywords: Predictive control, Control applications, Renewable Energy
Abstract: With rising global energy prices and the urgent need for carbon neutrality, enhancing energy efficiency has become a critical priority for railway operators. While regenerative braking significantly reduces consumption by sharing energy between accelerating and decelerating multiple trains, practical implementation is often hindered by the strict requirement for punctuality and the high computational cost of real-time optimization. This paper proposes a Multi-Horizon Model Predictive Control (MH-MPC) framework designed to achieve both precise motion control and high-accuracy adherence to arrival schedules. Unlike conventional single-horizon MPC, the proposed method integrates multiple horizons with varying temporal resolutions, enabling the controller to maintain punctuality over a long duration while delivering fine-grained inputs at each control cycle. We adopt a leader-follower coordination strategy where the follower's objective function is formulated to maximize the absorption of the leader's regenerative power. Numerical simulations demonstrate that the proposed MH-MPC approach achieves an energy reduction of approximately 6.5 % in the base scenario, which further increases to a maximum of 9.5 % when integrated with departure time optimization, all while strictly adhering to scheduled arrival times. Furthermore, recognizing that not all regenerative energy can be instantaneously utilized due to operational variations, we integrate a battery storage model into the system. Based on multi-scenario simulations, we present a systematic methodology for optimal battery sizing that prioritizes the Return on Investment (ROI). This study provides a practical engineering bridge between advanced control theory and cost-effective infrastructure investment for modern railway systems.

Keywords: Neural networks, Autonomous systems, Cooperative control
Abstract: This paper studies real time formation control for large scale multi agent systems (LMAS) with anonymous agents and finite time requirements. Instead of solving a centralized Hamilton Jacobi Bellman (HJB) problem or a coupled mean field game system, we design an assignment free controller directly at the distribution level. The swarm state is represented by a time varying probability density governed by a Fokker Planck equation with diffusion. We propose a Fokker Planck Neural Network (FPNN) to learn an admissible density evolution and a shared linear feedback law that satisfy the PDE residual, initial terminal constraints, and mass conservation on bounded domains. To quantify the performance tradeoff, we derive Riccati residual based certificates that upper bound the optimality loss relative to the centralized LQR benchmark. Simulations validate that the learned density evolution is consistent with classical PDE solvers, that the induced drift generates matching agent level trajectories for large populations, and that the overall computation cost is reduced by orders of magnitude compared with iterative mean field game solvers.

Keywords: Distributed control, Power systems, Smart grid
Abstract: This manuscript presents a novel distributed greedy framework applicable to a network of thermostatically controlled loads (TCLs) to desynchronize the network’s aggregated power consumption. Compared to the existing literature, our proposed framework offers two distinct novelties. First, our proposed algorithm relaxes the restrictive assumptions associated with the communication graph among TCLs. To elaborate, our algorithm only requires a connected graph to execute control, a condition less demanding than its counterpart algorithms that mandate a star architecture, K-regular graphs, or undirected connected graphs. Second, a significant novel feature is the relaxation of the obligation to share private information, such as each unit’s local power consumption and appliance temperatures, either with a central coordinator or neighboring TCLs. The findings presented in this brief are validated through simulations conducted over a network comprising 1000 TCLs.

Keywords: Fault-tolerant systems, Robotics applications, Aerospace applications
Abstract: This paper presents a nonlinear control law for a quadrotor unmanned aerial vehicle (UAV) under single-rotor failure that guarantees set stabilization via quasi-static feedback (QSF). Given a geometric curve in three-dimensional space, we characterize and stabilize the zero-dynamics manifold, also known as the path-following manifold, which represents all feasible motions along the path. Stabilizing this manifold ensures path-invariance: a UAV with a failed rotor initialized on the path with an appropriate orientation remains on the path for all future time. Furthermore, local exponential convergence to the manifold is guaranteed under certain conditions, implying that, under the stated assumptions, rotor failure during flight does not cause transverse deviation from the path. The proposed controller thus provides theoretical safety guarantees, which are validated through numerical experiments in the Drake physics-based simulation engine. The Code is publicly available at https://gradslab.github.io/quasistatic-ftc/

Keywords: Hybrid systems, Linear robust control, Robotics
Abstract: In robotic manipulation tasks involving contact, switching between force and position controllers can lead to undesired oscillations in contact forces and manipulator position. This paper presents an algorithm that enables a 6-link manipulator operating in three-dimensional space to robustly approach, grasp, and manipulate objects at predetermined locations under real-world noise and perturbations. The method incorporates hysteresis to mitigate chattering effects during contact with the object, thereby improving end effector stability and force regulation. Experimental results validate the effectiveness of the proposed algorithm through comparison with a discontinuous switching baseline, demonstrating improved robustness for robotic applications requiring stable manipulation.

Keywords: Process control, Control applications, Data analytics
Abstract: Industrial alarm systems are critical for the safe and reliable operation of multimode industrial processes, yet their performance is often degraded by static alarm configurations that ignore changes in operating modes. During mode transitions such as start-ups and shutdowns, this mismatch leads to alarms that are systematic consequences of the operating context rather than indicators of abnormal behavior, contributing to alarm flooding. This paper presents a context-aware, data-driven framework for mode-dependent alarm analytics that integrates unsupervised operating mode identification with transition-based alarm association analysis. Operating modes are inferred directly from continuous process measurements using clustering in a projected Linear Discriminant Analysis (LDA) space and temporally synchronized with Alarm and Event (A&E) logs to identify alarms consistently triggered during specific mode transitions. Mode-dependent alarm rules are then extracted on a unified timeline without requiring explicit mode labels or manual data segmentation. The proposed framework is validated using real operational data from an industrial pipeline booster station, demonstrating accurate reconstruction of valve-based operating modes and the identification of physically interpretable alarms associated with shutdown transitions, thereby providing a systematic basis for mode-dependent alarm rationalization.

Keywords: Process control, Fault detection/accomodation
Abstract: Alarm systems in industrial processes often suffer from alarm flooding, where excessive nuisance alarms overwhelm operators and compromise operational safety. This paper proposes a voting-based aggregation framework that systematically combines multiple alarm suppression modules to reduce false alarm rates while preserving fault detection capability. The main advantages of the proposed approach are its interpretability and configurability. In particular, the false-alarm and missing-alarm rates of the aggregated alarm system are analytically characterized, and optimal voting parameters are determined using a Bayesian log-likelihood ratio fusion rule. Finally, the effectiveness of the proposed method is validated on the Tennessee Eastman Process benchmark.

Keywords: Fault detection/accomodation, Fault-tolerant systems
Abstract: This paper presents a deployable multi-sensor anomaly detection pipeline for predictive maintenance of robot joint motors using synchronized temperature, voltage, and encoder position measurements. The proposed workflow performs preprocessing and temporal alignment, constructs lightweight temporally informed features (rolling statistics), and applies feature-level fusion prior to classification. Because confirmed fault annotations are often unavailable in practice, we generate proxy anomaly labels using interquartile range (IQR) fences on each sensor channel and fuse flags with an OR rule, yielding an anomaly prevalence of 26.12%. To reduce leakage from temporally correlated time-series data, we evaluate generalization using a session-based split across eight sessions and six motors. We compare three model classes: Random Forest, XGBoost, and an LSTM sequence model. We further integrate model outputs into a fault detection, isolation, and recovery (FDIR) framework that maps anomaly evidence to residual checks and staged recovery actions suitable for industrial supervisory control. The resulting implementation supports real-time deployment via a REST interface with a median single-prediction latency of 42 ms on a standard CPU platform.

Keywords: Cyberphysical systems, Machine learning, Reinforcement learning
Abstract: Cyber physical systems such as autonomous vehicles operate in highly dynamic environments where interactions with other autonomous and human agents is inevitable. Reinforcement learning (RL) is a well-established paradigm to allow agents to learn behaviors through interactions with the environment when a model of the environment is not known or available. When actions of any single agent in such multi-agent setups can be influenced by its own belief on other agents' actions and their risk-taking abilities, it becomes critical to quantify social preferences for each agent. In such a situation, it is also important to model widely observed preferences of human operators who will likely be sharing the same environment as autonomous agents. While significant progress has been made in modeling social preferences and risk-aware behavioral models into RL, these have largely occurred in parallel. This paper presents a two-pronged solution approach that uses cumulative prospect theory (CPT) to characterize risk-awareness of an individual agent and social value orientation (SVO) to represent the spectrum of egoistic to altruistic agent behaviors. We define the SVO-informed CPT-value of a random variable, and use it to design a novel subjective learning procedure that we term the CPT-SVO-MARL Algorithm. We empirically demonstrate stable learning behavior in a representative cyber physical control scenario.

Keywords: Learning, Cyberphysical systems, Computer-aided control design
Abstract: In this paper, we propose a method to reduce the uncertainty of learned over-approximations of monotone systems by optimizing the partition of the state space. We formulate the problem of finding a rectangular partition that minimizes the volume of the over-approximation on the entire state space, a measure of the conservatism in the data-driven reachability analysis. We introduce an Alternating Minimization (AM) algorithm that iteratively optimizes the partition along each dimension. A key contribution is the use of a Divide-and-Conquer Dynamic Programming approach for the 1D sub-problems, which computes the globally optimal partition in the scalar case and serves as the inner step of our multi-dimensional heuristic. We demonstrate the scalability and effectiveness of our approach through numerical examples, including high-dimensional datasets and a cruise control safety game, showing significant improvements over uniform partitioning.

Keywords: Neural networks, Sensor fusion, Robotics applications
Abstract: Surface electromyography (EMG) provides a non-invasive interface for decoding human motor intent and has been widely studied for rehabilitation and assistive robotics; however, reliable upper-limb motion intent classification remains challenging in rehabilitation scenarios, particularly for passive exoskeleton-driven movements where EMG activity is weaker and less discriminative than during voluntary motion. This paper proposes a lightweight con￾volutional neural network (CNN) framework for classifying upper-limb motions directly from raw EMG signals acquired during passive movement, targeting three representative re￾habilitation tasks: elbow flexion–extension, shoulder flexion– extension, and shoulder abduction–adduction. Six-channel surface EMG signals were acquired using the CLEVERarm upper-limb exoskeleton developed by our research team, segmented into overlapping windows, and evaluated using trial-wise cross-validation. Class imbalance due to differences in movement duration was addressed using a weighted loss function. Experimental results demonstrate that the proposed framework achieved a mean classification accuracy of 92.01% (±5.8%) and an F1-score of 92.01% across trials, with strong class discrimination measured by the area under the receiver operating characteristic curve (AUC: 0.97–0.98), while maintaining low computational complexity. The findings demonstrate that CNN-based decoding of raw EMG enables reliable motion classification during exoskeleton driven passive upper-limb movements, which is relevant for rehabilitation applications.

Keywords: Renewable Energy, Machine learning
Abstract: This paper presents an investigation of wind farm power maximization using the existing optimization method Serial-Refine, designed for wake steering in stationary layouts, applied to turbine repositioning in floating offshore wind farms. By commanding each turbine with a reference nacelle-yaw angle, the wake steering effect seen in stationary layouts leads to turbine repositioning and layout reconfiguration in floating offshore wind farms. We show that the Serial-Refine optimization method remains suitable for turbine repositioning, however, the reliance on simulation combined with the additional complexity of modeling the reconfigured layouts slows down the optimization. A proof-of-concept imitation learning approach is applied to reduce the optimization cost, whereby the optimal nacelle yaws produced by Serial-Refine are taken as expert behavior. By learning to mimic a Serial-Refine expert, we maintain optimization performance with reduced computation.

Keywords: Robust control, Reinforcement learning, Mechatronic systems
Abstract: Flexible motion systems are subject to model uncertainties and time-varying disturbances, which make high-precision, robust positioning challenging. This paper proposed a reinforcement learning (RL)-based method for the tuning of a Youla-parameterized in a nanopositioning flexible system. The framework was designed to enhance disturbance rejection and robustness under model uncertainties and time-varying disturbances. By embedding the learning process within the Youla structure, closed-loop stability was inherently preserved. An offline tabular Q-learning algorithm was developed, utilizing frequency-domain performance indices as state variables and incorporating robustness-aware reward shaping. Simulation results demonstrated performance improvements of 83.9% over the baseline controller and 35% over fixed Q designs.

Keywords: Predictive control, Aerospace applications
Abstract: Aero-engines operate under harsh and uncertain conditions, making safe and robust control challenging. To address this issue, we develop a robust model predictive control (RMPC) framework tailored for aero-engine applications. To further reduce the computational burden inherent in traditional RMPC, we propose an Efficient Tube MPC (ET-MPC) scheme that incorporates an event-triggered mechanism to limit optimization frequency while ensuring constraint satisfaction. The stability and feasibility of the method are established, and simulations on a real-world aero-engine model demonstrate that ET-MPC achieves safe, accurate tracking with significantly reduced computational cost.

Keywords: Predictive control, Ships and offshore vessels, Learning
Abstract: This paper presents a learning-based robust distributed model predictive control (DMPC) framework for heterogeneous autonomous surface vehicle (ASV) platoons subject to input bounds, matched disturbances, and inter-vehicle safety constraints. Each vessel solves a local, parallelizable MPC using exchanged assumed trajectories while enforcing a coupled minimum distance constraint. The control consists of a nominal tube-based model predictive control and an adaptive term produced online by a deep neural network that compensates the state-dependent uncertainties explicitly. A projection update and bounded activation layer guarantee a uniform bound on the network output, allowing it to be integrated into constraint tightening to ensure recursive feasibility and input-to-state (ISS) stability. Simulation results of a three-ASV platoon demonstrate safe spacing, accurate velocity tracking, and effective cancellation of matched disturbances, which confirms that embedding learning within a tube-based DMPC framework reduces conservativeness while retaining distributed scalability.

Keywords: Predictive control, Machine learning, Software tools
Abstract: Incorporating learning-based models, such as artificial neural networks or Gaussian processes, into model predictive control (MPC) strategies can significantly improve control performance and online adaptation capabilities for real-world applications. Still, enabling state-of-the-art implementations of learning-based models for MPC is complicated by the challenge of interfacing machine learning frameworks with real-time optimal control software. This work aims to fill this gap by incorporating external sensitivities in sequential quadratic programming (SQP) solvers for nonlinear optimal control. To this end, we provide L4acados, a general framework for incorporating Python-based dynamics models in the real-time optimal control software acados. By computing external sensitivities via a user-defined Python module, L4acados enables the implementation of MPC controllers with learning-based residual models in acados, while supporting parallelization of sensitivity computations when preparing the quadratic subproblems. We demonstrate significant speedups and superior scaling properties of L4acados compared to available software using a neural-network (NN)-based control example. Last, we provide an efficient and modular real-time implementation of Gaussian process-based MPC (GP-MPC) using L4acados, which is applied to two hardware examples: autonomous miniature racing, as well as motion control of a full-scale autonomous vehicle for an ISO lane change maneuver.

Keywords: Control applications, Predictive control, Mobile Robots
Abstract: Nonlinear model predictive control (NMPC) is widely used in robotics due to its ability to handle nonlinear dynamics and explicitly enforce state and input constraints by solving an optimal control problem online. However, standard NMPC implementations typically rely on direct time discretization and evaluate costs and constraints only at sampling instants. This paper investigates the real-world effectiveness of a previously introduced lifting-based NMPC formulation that explicitly accounts for intersample responses by minimizing a continuous-time tracking error criterion. The formulation is implemented for trajectory tracking of a differential-drive robot and evaluated against a conventional NMPC approach. Real-world experiments on lemniscate and right-triangular reference trajectories, under variations in target speed and sampling period, show that the lifting-based NMPC consistently achieves lower tracking RMSE and mitigates corner-cutting at vertices. Solve-time measurements further confirm that the approach can be executed online on embedded hardware while maintaining real-time feasibility.

Keywords: Control applications, Predictive control, Nonlinear robust control
Abstract: Time-minimal and collision-free motion of overhead cranes is critical in industrial applications with strict productivity and safety requirements. This paper presents a control framework that combines Artificial Reference Model Predictive Control (ARMPC) with a Higher-Order Control Barrier Function (HOCBF) to achieve near time-optimal point-to-point motion with explicit collision avoidance guarantees. The ARMPC formulation enables fixed-horizon optimization suitable for real-time deployment, while the HOCBF enforces safety by ensuring forward invariance of a collision-free set without significantly increasing computational complexity. The approach is tailored to a planar overhead crane system and validated through code generation, simulations, and experimental deployment on a laboratory-scale setup. A comparison with a free-time time-optimal control formulation demonstrates that the proposed method achieves near optimal transfer times with substantially improved real-time feasibility.