Keywords: Automotive Systems

Keywords: Assistive and Rehabilitation Robotics, Soft Robotics, Sensors and Actuators
Abstract: Reconfigurable compliant joints (RCJs) enable morphology-dependent compliance tuning, but this behavior has not yet been characterized when embedded within controlled actuator modules. This paper presents an experimental study of an alignment-tolerant actuator module in which interchangeable RCJ morphologies are integrated into a common position-controlled architecture to evaluate their effect on system-level response. With actuator command conditions and module hardware held constant, multiple RCJ morphologies were compared using torque-angle behavior, stiffness progression, video-derived central hub translation, and deformation at selected rotations. Aligned and imposed-misalignment conditions were further examined to assess passive accommodation within the module. The results show that RCJ morphology substantially influences module-level torque-angle response and stiffness development, while also altering deformation patterns and central hub translation. Under misaligned conditions, the tested morphologies exhibited distinct accommodation behaviors, indicating that the embedded RCJ acts as a morphology-dependent passive tuning element rather than solely as a compliant connector. These results establish the actuator module as a useful platform for evaluating RCJs at the system level and show that morphology can be used to shape both mechanical output and passive alignment accommodation. This work provides an experimental foundation for future module design, predictive morphology selection, and integration into broader assistive and alignment-sensitive actuation systems.

Keywords: Control of Smart Buildings and Microgrids, Control Applications, Agricultural Systems
Abstract: This paper proposes an Economic Nonlinear Model Predictive Control (ENMPC) framework incorporating Supplemental Artificial Lighting (SAL) for smart greenhouse production systems. SAL is explicitly incorporated into the nonlinear greenhouse model as a growth-driving control input that directly affects photosynthesis. A cumulative constraint is then introduced to enforce the minimum Daily Light Integral (DLI) requirement over the prediction horizon. Simulation results under extremely cloudy conditions show that the proposed method is able to ensure minimum DLI satisfaction and consequently improve the Economic Profit Index (EPI) compared to a conventional NMPC based on a classical greenhouse model.

Keywords: Manufacturing Systems, Control Applications
Abstract: Manufacturing accounts for nearly one-third of global electricity consumption, creating a strong need for energy-aware production. Such methods must improve energy performance without sacrificing production efficiency or practical executability. This requires an integrated modeling and control framework that captures both process-level physics, which determine energy consumption and feasible operation, and system-level dynamics, which govern throughput and production timing. This paper proposes a hierarchical System-Level Energy-Efficiency Digital Twin (SLEE-DT) framework for energy-aware production scheduling in batch manufacturing systems. The framework integrates modular DTs at the process and production-line levels to enable feasible, energy-aware decision-making. The hierarchical SLEE-DT consists of two decision layers: the upper layer solves an order scheduling problem under real-time electricity pricing, while the lower layer computes equipment-compliant process control inputs. The hierarchical SLEE-DT is experimentally validated on Iron MaBLe, a bench-scale automated beverage bottling line. Experimental results show that the hierarchical SLEE-DT accurately models production throughput and cycle time. Case studies of multiple-order scheduling scenarios demonstrate that the proposed approach reduces total energy costs relative to the flat-price baseline and heuristic demand-response baseline while satisfying all production deadlines.

Keywords: Mining, Mineral and Metal Processing, Manufacturing Systems, Estimation
Abstract: Metal peeling produces a thin metal strip or foil in a single step. Unlike rolling, strip thickness is not prescribed geometrically, but emerges from the coupled strip-formation mechanics and transport dynamics, making it sensitive to cutting conditions, tool wear, and feedstock variability. Existing control methods regulate transport variables, strip velocity or tension, to control thickness indirectly, which works under nominal conditions but rely on the degree to which strip-formation disturbances are reflected in the measured feedback. This article develops a direct thickness feedback architecture using a cascaded controller: an outer loop adjusts the strip-velocity reference based on measured thickness, and an inner loop regulates transport velocity via coiler torque. Two sensing configurations are considered: (1) thickness measurements at span entry with negligible delay, and (2) downstream measurement with a velocity-dependent delay. A variable-delay predictor is developed to reconstruct the current entry thickness from delayed measurements. A control-oriented formulation for the direct-thickness architecture is developed, which provides better rejection of disturbances, especially cutting-speed and depth-of-cut variations. The no-delay direct-thickness architecture gives the best overall performance, while the delayed-sensing architecture retains the same qualitative advantage but with reduced effectiveness due to transport delay and predictor mismatch during abrupt transients. The study establishes an efficient framework for direct real-time regulation of strip thickness in metal peeling and clarifies its advantages relative to indirect transport-based control.

Keywords: Transportation Systems, Control Applications, Intelligent Autonomous Vehicles
Abstract: This paper develops a hierarchical control framework for regulating mixed autonomous and human-driven traffic on freeway networks described by a multi-class METANET model. At the upper level, a multivariable Newton-based extremum-seeking (NES) controller determines class-specific density references, which are tracked at the lower level through variable speed limits using filtered feedback linearization (FFL). By estimating the local objective curvature online, the proposed NES approach reduces the dependence of the adaptation dynamics on the objective Hessian that characterizes conventional gradient-based extremum seeking. Local practical stability of the closed-loop hierarchical system is established under an explicit time-scale-separation condition. Simulations on a corridor modeled after I-485 demonstrate faster convergence, improved density regulation, and reduced upstream congestion spillback compared with gradient-based ES--FFL and PID-based control.

Keywords: Control Applications, Path Planning and Motion Control, Robotics
Abstract: Robotic autonomy in unstructured outdoor environments remains challenging due to uncertainty, sensor degradation, partial observability, and dynamic environmental variability. We present a closed-loop embodied AI framework for robust field robotics integrating multimodal perception, uncertaintyaware online adaptation, reinforcement learning (RL), and model predictive control (MPC). The proposed architecture consists of four tightly coupled modules: (i) an RGB-D perception encoder producing latent environmental representations, (ii) an uncertainty estimation module based on predictive entropy, (iii) an online self-supervised adaptation mechanism for representation refinement under distributional shifts, and (iv) a hybrid RL+MPC control stack in which RL generates nominal actions while MPC enforces safety and dynamic feasibility constraints. Experiments are conducted in NVIDIA Isaac Sim across randomized forestnavigation scenarios involving varying terrain slope, vegetation density, sensor noise, and partial occlusions. Compared with nonadaptive RL, MPC-only, and vision-only baselines, the proposed approach improves task success rate by 21% and reduces collision frequency by 35%. Ablation studies demonstrate that uncertaintytriggered adaptation and MPC safety filtering are both critical for robust autonomy in dynamic outdoor environments.

Keywords: Modeling and Validation, Cognition modeling, Multi-agent and Networked Systems
Abstract: This paper presents a wayfinding framework for pedestrian dynamic models of evacuation in which each agent continuously updates their routes on a predictive occupancy map. In the proposed framework, motivated by the cognitive concept of situational awareness, each agent uses anticipated motion to predict the future positions of nearby agents and construct a probabilistic occupancy field on a grid. Next, a path planning algorithm is used to build the shortest route to the exit on this occupancy map. Waypoints on the shortest route serve as intermediate goal locations within social-force dynamics. The model is validated by comparing macroscopic observables with experimental data of multiple groups ranging from three to nine individuals evacuating a square room. Beyond obtaining visually similar trajectories, statistical comparisons show that including wayfinding improves model performance in terms of matching average speed, turn rate, and nearest neighbor distance with real data. These results suggest possible directions to improve pedestrian evacuation dynamics by integrating cognitive concepts that enable wayfinding and path planning.

Keywords: Modelling, Identification and Signal Processing, Multi-agent and Networked Systems, Transportation Systems
Abstract: Identifying causal exogenous inputs is a key challenge in data-driven modeling of dynamical systems. We propose a dynamic mode decomposition with control (DMDc) based framework that infers such inputs based on their impact on reconstruction accuracy. A metric is introduced to evaluate whether including a candidate variable improves system estimation, providing a principled criterion for input selection. The approach is validated on canonical systems with controlled exogenous forcing and on synthetic traffic data with known interaction structure. Results show that the framework correctly identifies the immediate lead vehicle as causal while rejecting non-influential variables such as the rear vehicle, enabling inference of both presence and strength of exogenous influence.

Keywords: Multi-agent and Networked Systems
Abstract: The problem of fractional order text{PI}^beta text{D}^alpha consensus control of a double-integrator multi-agent system is considered assuming either an undirected or a directed communication network as well as Caputo fractional derivative and integral orders. The fractional control protocol provides additional flexibility in tuning the closed-loop response beyond integer order PID control. Necessary and sufficient stability bounds are obtained using Matignon's theorem in terms of the separate proportional, derivative, and integral gains and the fractional orders that guarantee global Mittag-Leffler stability of the consensus manifold.

Keywords: Multi-agent and Networked Systems, Large Scale Complex Systems, Intelligent Autonomous Vehicles
Abstract: Achieving robustness to disturbances while preserving scalability remains a fundamental challenge in large-scale heterogeneous leader–follower multi-agent systems with directed communication topologies. This paper develops a scalable robustness analysis framework that characterizes the influence of exogenous disturbances and leader-induced perturbations on consensus performance with network-size-independent guarantees. A scalable Input-to-Error Stability (sIES) formulation is introduced to quantify the worst-case leader-tracking error through an explicit L∞ input-to-error gain. Distributed matrix inequality conditions are derived that guarantee exponential convergence and scalable robust consensus based solely on local agent dynamics and uniform bounds on inter-agent coupling, without requiring global network information. The resulting conditions establish network-size-independent input-toerror performance guarantees for heterogeneous leader–follower networks over directed graphs containing a spanning tree. The proposed conditions further provide a unified interpretation of disturbance propagation that connects the classical notions of string stability and mesh stability within a scalable robustness framework. The proposed framework is validated through large-scale numerical simulations on heterogeneous networks with randomly generated communication topologies, demonstrating network-size-independent tracking performance over large heterogeneous networks. Practical applicability is further demonstrated on a Cooperative Adaptive Cruise Control (CACC) platform through distributed simulations and hardware-inthe- loop experiments integrating multiple F1Tenth autonomous vehicles with real-time virtual agents. Both numerical and experimental results confirm robust leader tracking, bounded spacing regulation, and scalable disturbance attenuation under persistent disturbances, thereby confirming the theoretical scalability and robustness guarantees.

Keywords: Multi-agent and Networked Systems, Path Planning and Motion Control, Control Design
Abstract: This paper studies cooperative multi-agent exploration in unknown environments with large 2-D free regions connected by narrow corridor passages. In such environments, conventional frontier-based exploration can underestimate narrow connectors, causing late traversal and ineffective coverage across different chambers. To address this issue, we propose a mixed-dimensional exploration framework in which chambers are explored using a coordinated 2-D frontier policy, while sufficiently narrow connectors are treated as transport links and handled separately by a commitment-based decision layer. The proposed controller combines frontier utility-cost assignment in 2-D regions with connector activation, traversal, and transition logic in the 1-D links, and alternates between them using a fixed priority hierarchy. We evaluate the proposed method in a representative mixed-dimensional chamber-corridor environment and benchmark it against an adapted Burgard frontier baseline and coExplore. Simulation results show that the proposed method achieves faster exploration progress, lower mean agent travel distance, and lower redundant overlap than the benchmark methods.

Keywords: Multi-agent and Networked Systems, Path Planning and Motion Control, Robotics
Abstract: Non-prehensile coordinated mobile manipulation is challenging because interactions between the robot bases and payload must be managed carefully. We present the Dynamic Admittance parameterization of Non-Prehensile Multi-robot Transport with Optimal Coordinated Planning (DYNAMO) architecture for payload transport using dual nonholonomic 4DOF mobile manipulators. DYNAMO minimizes snap and angular momentum change along a payload path and adapts arm motion to interaction forces. In hardware experiments DYNAMO achieves longer and more reliable payload transport than position-based methods, providing a practical foundation for multi-robot cooperative grasp-free manipulation.

Keywords: Mechatronic Systems, Robotics, Machine Learning in modeling, estimation, and control
Abstract: This paper presents a machine learning–based terrain detection system for bipedal locomotion using tactile sensing and inertial measurements. A lightweight terrain-sensing mechatronic system, using tactile pressure sensors and an inertial measurement unit (IMU), is designed and integrated into a bipedal robot to capture distributed foot-ground interactions. A long short-term memory (LSTM)-based neural network is developed to classify terrain types directly from segmented time-series sensor measurements without requiring calibration or handcrafted feature extraction. Experimental validation is conducted on a physical bipedal robot across six representative terrain conditions. The proposed method achieves an overall classification accuracy of 93.75% and consistently outperforms approaches using single-modality sensing. The results demonstrate the effectiveness of tactile-informed learning for robust terrain recognition and provide a foundation for terrain-aware locomotion control.

Keywords: Human-Machine and Human-Robot Systems, Machine Learning in modeling, estimation, and control, Estimation
Abstract: Accurate prediction of human motion using pose information for predicting human trajectories is essential in applications in collaborative environments involving human–robot or human-machine interaction, enabling proactive collision avoidance and a safer workplace. However, forecasting realistic human motion remains challenging due to the high dimensionality, nonlinear dynamics, and the noisy nature of human motion. In this work, we propose a encoder-decoder based Conditional Generative Adversarial Network (cGAN) framework for human motion prediction that generated future pose trajectories from their respective historical observations. The framework employs Feature-wise Linear Modulation (FiLM) to condition both the generator and discriminator on action semantics, enabling action-specific motion generation. The generator learns a distribution over plausible motion trajectories while ensuring spatial and temporal realism. To improve the realism and physical plausibility of the generated trajectories, the training objective incorporates the skeletal constraints together with frequency-aware regularization, preserving both kinematic consistency and the natural temporal rhythms of human motion. An auxiliary Residual Refinement Network (RRN) further reinforces the output by capturing complex inter-limb dependencies. Experimental evaluation on the NTU RGB+D dataset demonstrates that the proposed framework generates smooth, physically consistent, and temporally coherent motion trajectories while maintaining real-time inference capability, making it suitable for safety-critical human–robot collaboration and digital twin applications.

Keywords: Machine Learning in modeling, estimation, and control, Optimal Control, Nonlinear Control Systems
Abstract: We develop a physics-informed learning framework for energy-shaping control of port-Hamiltonian (pH) systems from trajectory data. The proposed approach co-learns a pH system model and an optimal energy-balancing passivity-based controller (EB-PBC) through alternating optimization with policy-aware data collection. At each iteration, the system model is refined using trajectory data collected under the current control policy, and the controller is re-optimized on the updated model. Both components are parameterized by neural networks that embed the pH dynamics and EB-PBC structure, ensuring interpretability in terms of energy interactions. The learned controller renders the closed-loop system inherently passive and provably stable, and exploits passive plant dynamics without canceling the natural potential. A dissipation regularization enforces strict energy decay during training, thereby enhancing robustness to sim-to-real gaps. The proposed framework is validated on state-regulation and swing-up tasks for planar and torsional pendulum systems.

Keywords: Control Applications, Optimal Control, Machine Learning in modeling, estimation, and control
Abstract: The use of reinforcement learning policy iterations for a hydrogen gas processing system that supplies hydrogen for fuel cells is presented. In the considered problem, the produced fuel cell current plays a role of a disturbance to the hydrogen processing system so that the goal of optimization is to reduce the impact of the fuel cell produced electric current on the hydrogen processing system. This problem can be formulated as a zero-sum differential game. The presented results show that the optimal disturbance suppression can be achieved in only three or four reinforcement learning interactions with the hydrogen gas processing system, during which only the standard algebraic Riccati equations need to be solved iteratively. This is much simpler than solving the zero-sum differential game algebraic Riccati equation. Moreover, we have simplified derivations used for solving the zero-sum differential game algebraic Riccati equation in terms of standard algebraic Riccati iterations.

Keywords: Robotics, Machine Learning in modeling, estimation, and control, Control Design
Abstract: Kinematics-driven underactuated balance robots, such as autonomous bicycles and Segways, require the coordination of trajectory tracking and balance stabilization under limited controls. When system dynamics are unknown, reinforcement learning (RL) designs are computationally expensive and struggle to deal with dynamic coupling and underactuation. This paper proposes a control method integrating RL with the robot's dynamic properties. The overall control is decoupled into trajectory control and balance control. As a result, the original control problem is reduced to an RL tracking problem for a 1-degree-of-freedom system. The GP model predicts the balance equilibrium online without requiring dynamics inversion. Simulations on an autonomous bicycle robot validate that the proposed approach achieves effective balance stabilization and accurate trajectory tracking in various unseen scenarios.

Keywords: Sensors and Actuators, Machine Learning in modeling, estimation, and control, Estimation
Abstract: Low-cost Micro-Electro-Mechanical Systems (MEMS) Inertial Measurement Units (IMUs) are widely used in robotics, aerospace, and autonomous systems. Yet, they suffer from significant angular errors, including bias instability, drift, and noise. Traditional calibration methods often rely on expensive external systems such as turntables or motion capture setups, limiting their applicability in embedded or field-deployable platforms. This paper presents a novel calibration framework that leverages a 6-degree-of-freedom (DOF) robotic arm and encoder-supervised deep learning to correct angular errors in IMU gyroscope measurements, specifically the Z-axis (Yaw). A Pixhawk 2.1 IMU was mounted on the end-effector of a MyCobot 320 Pi robotic arm to generate consistent and repeatable motion. Ground-truth orientation was estimated from encoder-derived joint angles and end-effector trajectories, then fused via a complementary filter to produce a robust supervision signal. A hybrid Convolutional Neural Network and Bidirectional Long Short-Term Memory (CNN-BiLSTM) model was developed to learn and compensate for gyroscope drift and bias using raw IMU data and a fused orientation reference. The proposed model was trained and evaluated across both static and dynamic datasets, achieving high accuracy with a Root Mean Squared Error (RMSE) below 0.1°/s and an R² score above 0.98. Three supervision schemes—encoder-only, end-effector-only, and fused—were compared, with the fused model consistently demonstrating superior performance and generalization. The system operates in real time and is compatible with lightweight embedded platforms, offering a scalable solution for in-field IMU calibration that does not depend on external references. This work establishes a pathway for deploying real-time, data-driven calibration in robotic, aerial, and space-based systems, significantly enhancing orientation estimation accuracy and long-term sensor reliability in GPS-denied or autonomous environments.

Keywords: Distributed Parameter Systems, Estimation, Power and Energy Systems
Abstract: The state of charge (SOC) in rechargeable batteries is a function of spatial position inside the battery. Typically, observers attempt to estimate either the SOC at the electrolyte boundary or the averaged SOC over the domain (bulk SOC). In demanding but realistic battery operation, bulk SOC can be an insufficient approximation of the distributed SOC, leading to negative efficiency and safety consequences. In this paper, the bulk SOC is used as an input measurement to estimate the distributed SOC, using an existing partial differential equation (PDE) model for cathode dynamics and the PDE backstepping technique, which drives the observer error toward the dynamics of a desired target system via a Volterra integral transformation. The invertible transformation is fully defined, and a proof of exponential stability for the target system is provided.

Keywords: Machine Learning in modeling, estimation, and control, Estimation, Power and Energy Systems
Abstract: We present a hybrid, physics-informed battery state-of-charge (SOC) estimator with state-of-health (SOH)-coupled parameter adaptation. The series resistance is identified online by recursive least squares, matching a model-free DCIR measurement to 0.1%; an unscented Kalman filter (UKF) on a 1RC model provides the physics baseline; and a BiLSTM trained exclusively on the UKF residual returns a hard-limited (±5%) correction the network cannot overrule. All results use open data with the held-out axis of every split stated. The residual pays off precisely where the physics is observability-limited: on the LiFePO4 flat OCV plateau, where any well-tuned filter is stuck near 9%, it cuts SOC RMSE to 1.67% (held-out temperature); on sloped-OCV NMC it refines 1.6% to 1.3% (held-out cell, wrong 80% init), and the physics backbone alone recovers a wrong init to 6.0% where coulomb counting (19%) cannot. SOH from incremental-capacity (dQ/dV) features reaches 0.9/1.2/1.7% RMSE (LiFePO4/NCA/NMC, leave-one-cell-out), and the SOH→capacity loop, driven end-to-end by the estimated SOH on held-out Oxford cells, cuts aging-induced SOC drift from 6.6% to 0.75%. Hardware and edge deployment appear in an extended version.

Keywords: Machine Learning in modeling, estimation, and control, Estimation, Power and Energy Systems
Abstract: Accurate state-of-charge (SOC) estimation is a key function of battery management systems. However, SOC is not directly measurable, and the terminal voltage is jointly shaped by nonlinear RC polarization dynamics and a temperature-dependent open-circuit voltage (OCV)-SOC relationship. Purely data-driven neural methods such as LSTMs can fit SOC well but often generalize poorly to unseen operating conditions. Classical Koopman-based methods lift the nonlinear dynamics into an approximately linear latent space, but require hand-crafted observable functions. Deep Koopman methods replace the manual observables with neural encoders, yet their linear latent dynamics remain insufficient for strongly nonlinear battery behavior. To address these issues, we propose a Deep Koopman-style framework that learns the lifting map from voltage, current, and temperature measurements through a gated recurrent unit (GRU) encoder. To further improve the generalization under temperature variations, we introduce an MLP-based latent dynamics module inspired by joint-embedding predictive architecture (JEPA) world models. Experiments on the LG HG2 dataset show that our proposed method achieves better SOC accuracy across temperatures and better generalization to unseen temperatures. Principal component analysis (PCA) further shows that our method learns a more generalizable encoder representation across temperatures.

Keywords: Power and Energy Systems
Abstract: This paper optimizes the rate of freshwater production by a wave glider-based reverse osmosis system. A wave glider is a mobile system that uses renewable wave energy for propulsion, and can also be retrofitted with a reverse osmosis subsystem. Thanks to their mobility, wave gliders are potentially well-suited for coastal water desalination in the aftermath of natural disasters. The literature already models and optimizes wave glider dynamics for surge performance, but the problem of optimizing wave glider-based desalination remains relatively less explored. We address this gap in the literature through a “what if” dynamic model-based design optimization study. The study focuses on a wave glider system sized to accommodate 3,800 liters of freshwater. We equip this system with a heavy submerged reaction mass to increase the force applied to the desalination piston pump. Moreover, we use a nonlinear return spring to allow this pump to achieve a larger stroke while producing output pressures suitable for reverse osmosis. Finally, assuming a variable geometry piston pump, we optimize the pump’s effective cross-sectional area plus the scaling of its return spring to maximize freshwater production for a stochastic portfolio of incoming wave conditions. This furnishes an appealing mean freshwater production rate above 900 liters per hour, plus a very large pump stroke per wave, intuitively analogous to the operation of traditional vibration-based energy harvesters.

Keywords: Power and Energy Systems, Estimation, Modelling, Identification and Signal Processing
Abstract: Accurate state-of-charge (SoC) estimation is essential for lithium-ion battery management systems, yet SoC cannot be measured directly in real time and must instead be inferred from terminal voltage and current. This paper presents a design for adaptive parameter identification and nonlinear-observer-based SoC estimation using a 2-RC equivalent circuit model. A dynamic regressor extension and mixing (DREM) algorithm is used to identify the physical model parameters online from voltage-current measurements without direct SoC information, while preserving the SoC integrator dynamics and providing an accurate observer-side model. The identified parameters are then used to construct a Jacobian-weighted nonlinear observer with a constant Lyapunov-based gain. Simulation results over a 20-hour combined charge-discharge cycle show that adaptive identification improves model consistency and enables faster convergence and more accurate SoC tracking. Compared with a baseline extended Kalman filter (EKF) under parametric uncertainty, the proposed nonlinear observer maintains nearly unchanged post-transient performance as mismatch increases. These results show that the proposed design provides a robust and computationally efficient solution for battery management applications.

Keywords: Adaptive and Learning Systems, Automotive Systems, Estimation
Abstract: Lithium-ion batteries often experience considerable heat generation during charge and discharge operations. The boundary cooling coupled with non-uniform heat generation leads to pronounced spatial temperature gradients which are computationally burdensome to estimate online. The lumped thermal assumption can compromise model accuracy, and the effectiveness of thermal control strategies. The paper produces a linear parameter varying thermal model by modeling the cell nodes' thermal behavior through analysis of heat generation and transfer mechanisms, which is then used as a boundary condition for a finite-difference heat balance model of the cell. A linear adaptive observer is implemented to estimate system states while accounting for unmeasured disturbances. The developed observer is robust to bounded parametric uncertainties and validation under multiple cycling conditions shows the thermal gradient prediction error to be less than 1C with a computation time below 0.05 ms/step, demonstrating suitability for real-time battery management applications.

Keywords: Adaptive and Learning Systems, Unmanned Ground and Aerial Vehicles, Robotics
Abstract: Complete loss of two opposing rotors leaves a quadrotor vulnerable to false-data injection attacks (FDIAs) in its virtual pseudo-control channels. We integrate an adaptive resilient control Lyapunov function quadratic program (RCLF--QP) with a model-based extended state observer (MB--ESO): the observer rejects lumped aerodynamic disturbances, while the QP modifies the nominal pseudo-control to enforce resilient CLF inequalities. For FDIAs with at-most-exponential growth, the altitude and reduced-attitude tracking errors are uniformly ultimately bounded. At 500,mathrm{Hz}, fixed-step simulations show bounded tracking, whereas the MB--ESO baseline loses performance.

Keywords: Control Design, Unmanned Ground and Aerial Vehicles, Aerospace
Abstract: A tethered autogyro with articulated rotors can operate as an unmanned aerial vehicle capable of energy-efficient, long-duration deployment by utilizing ambient wind energy to sustain flight. This article presents a model-based attitude control technique for such a system using a full-fidelity dynamic model. Using Lagrangian approach combined with Blade Element Momentum Theory and catenary mechanics, a previously developed 2D hybrid model is extended to three dimensions. The new model describes the complete rigid-body motion of the frame, including roll and yaw dynamics, and is augmented with rotor speed and flapping degrees of freedom for each blade. Equilibrium characteristics are examined through steady-state responses and compared with the prior model. The resulting trends of equilibria are found to be consistent with those reported in the literature. A feedback control strategy based on regenerative differential rotor braking is developed to modulate all three attitude angles. Simulations demonstrate effective attitude regulation and stable flight.

Keywords: Optimal Control, Machine Learning in modeling, estimation, and control, Unmanned Ground and Aerial Vehicles
Abstract: This paper studies online transfer for low-level quadrotor attitude-and-rate control under post-deployment dynamics variation. We separate post-deployment variation into a structured component that can be parameterized before deployment and a residual component that may only be identified online. A parameterized approximate optimal controller is constructed as a reusable base policy, while a bounded online residual layer is invoked only when the transferred dynamics are not adequately represented by the base parameterization. A sufficient condition is also provided under which residual adaptation is unnecessary. The proposed approach is examined in a high-fidelity nonlinear quadrotor simulator under representative actuator and inertia variation.

Keywords: Path Planning and Motion Control, Multi-agent and Networked Systems, Unmanned Ground and Aerial Vehicles
Abstract: This paper presents a methodology for adaptive multi-agent patrolling of a region using concentric circular orbits. Using the proposed line-of-sight kinematics, agents initially converge to a nominal orbit and maintain continuous circumnavigation to provide persistent coverage. Upon a reconfiguration requirement, agents adapt their formation across inner, nominal, and outer orbits to respond accordingly. The framework also includes re-arrangement of all agents back to the original orbit as normal conditions return. Stability analysis of the radial kinematics ascertains desired agent behavior for arbitrary initial conditions. In addition, the proposed approach provides bounded, closed-form velocity inputs for each agent. Simulation results demonstrate the effectiveness of the approach in representative patrolling scenarios.

Keywords: Unmanned Ground and Aerial Vehicles, Machine Learning in modeling, estimation, and control
Abstract: Real-world deployment of acoustic anomaly detection for identifying motor or propeller failures in Unmanned Aerial Systems (UAS) requires robustness to dynamic operating conditions. This paper presents an empirical study based on an original experimental dataset of Brushless DC (BLDC) motor acoustic signatures collected across nine distinct operational maneuvers. A semi-supervised 1D Convolutional Autoencoder is used to learn the healthy operational manifold from unlabeled data. The proposed approach achieves approximately 97% precision and 92% recall, demonstrating that reliable acoustic health monitoring can be achieved without requiring labeled fault data. These results support the integration of real-time acoustic diagnostics into In-time Aviation Safety Management Systems (IASMS).

Keywords: Unmanned Ground and Aerial Vehicles, Motion and Vibration Control, Distributed Parameter Systems
Abstract: Tethered aerial robots provide continuous power supply and extended operational endurance, but their dynamics become significantly more complex when the tether is flexible and the ground vehicle is in motion. This paper presents a dynamic model and structure-preserving simulation framework for a tethered drone attached to a moving ground vehicle through an elastic tether and winch mechanism. The system is derived using an extended Hamilton's principle, leading to coupled equations of motion for the tether, drone attitude, winch, and base motion. To accurately capture the nonlinear coupled dynamics while preserving geometric structure, a Lie group variational integrator is developed for the discretized system. The model is then used to study a hovering drone with a fixed base and with prescribed base vibrations representative of off-road vehicle motion. The results show that base excitation can induce significant tether oscillations and drone position deviations, especially near the tether's modal frequencies, while a flexible-tether model captures important behaviors that are not represented by a straight-tether approximation. The framework provides a useful tool for analyzing and designing control strategies for tethered drone systems operating with mobile ground vehicle platforms.

Keywords: Robotics, Manufacturing Systems

Keywords: Uncertain Systems and Robust Control, Control Design, Estimation
Abstract: This paper is motivated by the problem of controlling dynamic systems for safety. The literature typically pursues this goal using control barrier function (CBF) methods. These methods provide rigorous safety guarantees when they successfully ensure the forward invariance of user-defined safe sets. Unfortunately, exogenous disturbances and unmodeled dynamics can make it difficult to ensure safety using model-based CBF methods. Motivated by this challenge, existing work in the literature combines CBF control with disturbance estimation in pursuit of robust safety. Specifically, existing research uses disturbance observers to estimate errors in the model-based computation of CBF constraints. This makes it possible to incorporate the resulting disturbance estimates in CBF-based safety control. The goal of this paper is to present and compare the use of two different disturbance estimation algorithms for CBF-based safety control, namely: recursive least squares estimation and interval estimation. The paper presents both of these approaches, then compares them using examples focusing on battery thermal management and safe biomedical perfusion. Safety robustness improves with both techniques, particularly the interval estimation-based technique.

Keywords: Machine Learning in modeling, estimation, and control, Nonlinear Control Systems, Uncertain Systems and Robust Control
Abstract: Learning-based models are increasingly used to describe complex nonlinear systems, yet their deployment in safety-critical tasks is limited by modeling uncertainty, requiring a deeper understanding of the system's future behavior under modeling uncertainty. This paper presents a data-driven framework for a confidence-aware robust forward reachability of nonlinear systems with learned dynamics. The approach combines machine learning-based representation of the system dynamics with conditional conformal prediction to construct finite-sample, state-dependent probabilistic uncertainty bounds on model mismatch without assuming an uncertainty distribution. These uncertainty bounds are incorporated into a recursive reachable set propagation scheme, yielding robust forward reachable sets that contain the true system trajectory with a prescribed confidence level over a finite horizon, even under model misspecification. To enable practical use in downstream tasks, e.g., collision checking, the resulting reachable sets are represented using alpha shapes to capture nonconvex geometry and then decomposed into collections of convex polytopes for efficiency. The proposed method is demonstrated in numerical simulations on a Wave Adaptive Modular Vessel (WAM-V) unmanned surface vehicle using a Koopman operator-based representation of the system dynamics, where results demonstrate the efficiency and accuracy of our approach.

Keywords: Nonlinear Control Systems, Uncertain Systems and Robust Control
Abstract: This paper develops an implicit Euler discretization framework for the exponential fixed-time stable controllers of discrete-time nonlinear systems, which involve power-law and Lambert W nonlinearities. We show that the generalized equation defining the implicit control admits a unique solution at each time step by virtue of strict monotonicity, so that the controller is well defined without the introduction of additional parameters. Since the state-dependent gain vanishes at the origin, the implicit scheme yields exponential rather than exact fixed-time convergence. We formalize this behavior through the notion of emph{exponential fixed-time stability}, wherein the settling-time bound is independent of the initial conditions and decomposes into a transition phase and a halving phase. We further establish that, in the presence of bounded disturbances, the closed-loop system is ultimately bounded, with a residual set determined by the inverse of the gain. Finally, a spacecraft attitude control example is provided to illustrate the reduction in control effort afforded by the implicit scheme.

Keywords: Nonlinear Control Systems, Uncertain Systems and Robust Control, Cyber physical systems
Abstract: This letter introduces a new notion of Self-Scaling Attack-Resilient Control Lyapunov Functions (SSAR-CLFs) for nonlinear affine systems subject to a broad class of cyberattacks, including polynomially and exponentially unbounded false-data injection (FDI) signals acting on the control input. In contrast to the conventional ISS-based CLF methods and recent robust or fault-tolerant control frameworks that rely on bounded disturbance, or fault assumptions, the proposed approach guarantees stability under such attacks. This is achieved through a self-scaling gain mechanism that dynamically enforces the Lyapunov decrease condition online. A quadratic program (QP)-based controller is developed to implement the proposed SSAR-CLF condition in real time. Using comparison-based analysis, it is shown that the adaptation gain grows sufficiently fast to counteract even exponentially unbounded attack signals, so that the Lyapunov function becomes nonincreasing outside a compact set after a finite time. Consequently, the closed loop system is guaranteed to be uniformly ultimately bounded. Numerical results demonstrate the effectiveness of the proposed method compared with existing robust and ISS-based CLF-QP approaches.

Keywords: Adaptive and Learning Systems, Uncertain Systems and Robust Control, Control Applications
Abstract: Data-Enabled Predictive Control (DeePC) is an established model-free approach to predictive control, but two practical obstacles limit its deployment: computational cost that grows cubically with dataset size, and performance degradation when the data are corrupted. This paper introduces Robust Data Selection DeePC (RDS-DeePC), which treats both problems through influence function analysis. We derive a sensitivity score quantifying the leverage each trajectory segment exerts on the optimization solution, and we show that high-sensitivity segments typically correspond to outliers whereas low-sensitivity segments represent consistent data. Keeping only low-sensitivity segments therefore reduces computation and filters outliers in a single step, with no need for data quality labels. A two-stage online selection scheme, accelerated by the LiSSA algorithm, extends the framework to nonlinear systems. In simulation on four benchmark systems (an LTI DC motor, an inverted pendulum, a planar quadrotor UAV tracking a figure-8 trajectory, and a kinematic bicycle vehicle following a figure-8 path), RDS-DeePC selects 93 to 97% clean data and delivers comparable or better tracking performance under 20% data corruption.

Keywords: Uncertain Systems and Robust Control, Stochastic Systems, Nonlinear Control Systems
Abstract: Shared control in semi-automated vehicles requires dynamic, bidirectional arbitration between human and automation. A central challenge is maintaining safety while reducing unnecessary opposition despite uncertainty in driver intent and interaction dynamics. This paper presents a learning-based predictive arbitration framework that integrates Gaussian process regression with multistage model predictive control. Human-in-the-loop driving simulator data are used to learn residual mismatch between predicted and observed shared-control behavior. The learned probabilistic model is embedded within a scenario-based MPC to adapt automation impedance and authority online. Closed-loop obstacle-avoidance simulations show that learning resolves conflict more effectively than a worst-case baseline by preserving safety, reducing conservatism and modeled resisting interaction, and producing smoother automation-impedance transitions.

Keywords: Control Applications, Manufacturing Systems, Robotics
Abstract: In robotic machining, precision motion control involves tracking within the working regions and transitions between different regions. Achieving high-precision machining is challenging as vibrations can be induced after the transition, degrading the machining quality, while conventional feedback-based controlis fundamentally limited by the performance–robustness trade-off. The main contribution of this paper is the development of a data-driven tracking-transition approach that integrates the optimal output transition design with an iterative learning control-based feedforward-feedback scheme. The transition trajectory is designed to ensure smooth motion while minimizing output vibration. In addition, a data-driven robust optimal iterative learning control method is combined with computed torque control to construct a feedforward-feedback controller to accurately track the desired tracking-transition trajectory. The proposed method is validated through robotic machining trajectory tracking in the experiment. Experimental results demonstrate that the proposed approach outperforms conventional methods in achieving precision tracking-transition control for robotic machining.

Keywords: Machine Learning in modeling, estimation, and control, Estimation, Cyber physical systems
Abstract: : Gearbox condition diagnosis using vibration signal analysis leverages the correlation between gear condition and vibrations induced by the periodic meshing of gears. However, these vibrations are also dependent on operating speed, ambient noise, transmission path, and other factors. To highlight the periodic gear meshing features among these factors, synchronous averaging has been proposed as a model-agnostic physics-grounded preprocessing method that can enhance key signal features corresponding to gear condition. Nevertheless, objective evaluations regarding the impact of different synchronous averaging techniques are scarce, with evaluations limited to single domain applications. The objective of this paper is to compare the advantages and disadvantages of two synchronous averaging techniques with respect to clustering and machine learning (ML)-based gearbox condition diagnosis. Building upon our previous work, vibration data gathered from a physical system is preprocessed using time-synchronous averaging and angle-frequency domain synchronous averaging to highlight periodic features. The synchronously averaged data is then evaluated using statistical clustering and dimensionality reduction clustering methods, as well as 1D and 2D ML models to characterize the impacts of different preprocessing methods on gearbox condition diagnosis. The case studies show that certain preprocessing methods are better suited to certain ML applications, and clustering methods yield similar outcomes.

Keywords: Control Design, Marine Systems, Power and Energy Systems
Abstract: This work presents the numerical characterization of the Power Take-Off (PTO) system for a tunable laboratory-scale two-body Point Absorber (PA) Wave Energy Converter (WEC). Three PTO architectures are developed and compared using time-domain simulations in WEC-Sim under regular wave conditions: an ideal Spring–Damper (SD) baseline, a Direct-Drive Linear Generator (DDLG), and an Electrical Generator Equivalent Circuit (EGEC). Under quasi-static electrical assumptions, both electrical systems can be parameterized as a tunable linear damper, whose coefficient is set to match the optimal damping target. The DDLG is tuned through its electrical load resistance, while the EGEC offers an additional tuning parameter through its rack-and-pinion transmission ratio. Our study demonstrates deficiencies in the existing WEC-Sim model, and applies efficiency corrections to obtain experimentally consistent power estimates. The results establish a validated numerical baseline for the forthcoming experimental campaign at the University of New Hampshire (UNH) wave tank, with predicted mechanical power outputs reaching up to 37 [W] and corrected electrical outputs of approximately 26 [W] and 21 [W] for the DDLG and EGEC, respectively, at a wave frequency of ω= 4.0 [rad/s] and wave amplitude of η= 0.16 [m].

Keywords: Control Design, Marine Systems, Optimal Control
Abstract: This paper presents a hardware-in-the-loop (HIL) testbed for a belt-driven oscillating surge wave energy converter (OSWEC), enabling dry-lab experimental evaluation of power-takeoff (PTO) control strategies. The proposed platform combines a real-time dynamic emulator with a physical PTO system, allowing the OSWEC hydrodynamic response to be reproduced while preserving the hardware-level power generation process. A position tracking controller is developed for real-time motion reproduction, achieving a normalized root-mean-square error (NRMSE) of 6.82% and a root-mean-square tracking error of 0.527◦ over the 0.2–0.63 Hz frequency range and ±30◦ motion amplitude range. To demonstrate the capability of the testbed, an impedance-matching-based PI control strategy is implemented for the PTO system. Results under 0.25 Hz and 0.5 Hz regular-wave conditions (about 12.6 s and 6.3 s full-scale periods) show that the platform can evaluate PTO control performance in the presence of practical hardware non-idealities, and that the electrical power is more sensitive to the proportional gain than to the integral gain.

Keywords: Sensors and Actuators, Control Design, Mechatronic Systems
Abstract: The target of this research is to develop an absolute displacement sensor measurable mechanical vibration up to 1 meter amplitude and frequency as low as 0.1 [Hz] by applying active control to a proof-mass type seismometer. In previous studies, feedback (FB) gain is derived by using LQ feedback while feedforward (FF) gain applied to absolute acceleration is tuned experimentally by using classical control. In this study, FB and FF gains are derived as single FB gain by using LQ control with state of equation to which pseudo relative acceleration is taken into account. Besides, signal processing using a phase delay compensator is also applied to modify frequency characteristics. The designed sensor is evaluated through control experiments. According to these improvements, the dynamic range of the sensor is extended to 0.12~8.0 [Hz], while expected measurable vibration amplitude reached 0.7 meter.

Keywords: Machine Learning in modeling, estimation, and control, Estimation, Electromechanical systems
Abstract: Electrical impedance tomography (EIT) reconstructs the internal conductivity distribution of a body from boundary voltage measurements, but the associated inverse problem is severely ill-posed and computationally intensive. We propose a Bayesian physics-informed deep operator network (DeepONet) framework that learns a continuous mapping from voltage measurements to conductivity fields. A branch network encodes the measurement vector and a trunk network encodes spatial query coordinates; their inner product yields pointwise conductivity estimates at any location in the domain. A physics-informed positional encoding is introduced in the trunk network to embed the radial sensitivity structure of EIT directly into the spatial basis functions. The branch network is further extended to a Bayesian formulation that produces spatially resolved predictive uncertainty alongside the mean reconstruction, with uncertainty concentrating at inclusion boundaries where the inverse problem is most underdetermined. Training with randomly sampled query points and a weighted loss function that emphasizes anomaly regions ensures generalization and sensitivity to resistive or conductive inclusions. The resulting operator supports zero-shot super-resolution and real-time inference without retraining when the evaluation grid changes.

Keywords: Multi-agent and Networked Systems, Control Design, Nonlinear Control Systems
Abstract: This paper studies the multi-agent leader-follower formation tracking problem of pose and velocities for a network of N rigid body vehicles in which the dynamics of each agent evolves on TSE(3), the tangent bundle of the special Euclidean group SE(3). The objective is to drive a subset of leader agents to a prescribed reference rigid-body motion while simultaneously regulating the relative poses of the remaining agents so that the entire network achieves a desired moving formation. In contrast to approaches based on exponential coordinates or separate attitude and translational designs, the proposed method is formulated directly on (SE(3)) using group pose errors and rigid-body velocities. A geometric control law is constructed from leader tracking terms and follower relative-formation terms, preserving the coupled rotational and translational structure of the rigid-body dynamics.

Keywords: Multi-agent and Networked Systems, Estimation, Path Planning and Motion Control
Abstract: Cooperative multi-agent networks deployed for target localization and tracking remain critically vulnerable to malicious cyber attacks, since a single compromised agent can corrupt the centralized target belief and mislead the estimation process across the entire network. This paper presents a resilient multi-agent target localization and tracking framework under a limited agent sensing range and false-data injection (FDI) attacks. The proposed method combines information-based navigation for target search and posterior target state uncertainty reduction, Bayesian attack graph (BAG) based attack detection for probabilistic identification of the compromised agents, and reachable set-guided recovery that guides the uncompromised agents to relocalize the target when the sole localizing agent is attacked. The framework preserves the network’s localization capability after an attack by trust-gating compromised sensing information and guiding attack-immune agents toward a dynamically constructed recovery set representing the new target belief using a Linear Quadratic Regulator (LQR). Simulation validates that the proposed framework maintains a robust tracking performance under FDI attacks, demonstrating the potential for resilient autonomy in safety critical applications.

Keywords: Optimal Control, Intelligent Autonomous Vehicles, Multi-agent and Networked Systems
Abstract: Computational complexity has been a major challenge in game-theoretic model predictive control (GT-MPC), as real-time solutions to a game (e.g., Nash equilibria (NEs)) have to be computed at each sampling instant of an MPC. This challenge is especially critical in autonomous driving, where interactions may involve many agents, and decisions must be made at fast sampling rates. We show that this challenge can be addressed through time-distributed solution-seeking iterations designed based on Newton and Newton--Kantorovich methods. Specifically, the autonomous vehicle (AV) decision-making problem is formulated as a GT-MPC problem. To ensure solution attainability, a potential game framework is adopted. Within this framework, both potential-function optimization and best-response dynamics are used to seek the NE. To enable real-time implementation, Newton and Newton--Kantorovich methods are employed to solve the optimization problems arising in the NE-seeking algorithms, with their iterations distributed over time. Numerical experiments on an intersection-crossing scenario demonstrate that the proposed methods achieve effective real-time performance.

Keywords: Path Planning and Motion Control, Multi-agent and Networked Systems, Unmanned Ground and Aerial Vehicles
Abstract: This paper presents a decentralized framework for path following guidance of multiple uncrewed aerial vehicles (UAVs), while maintaining uniform spacing along a reference path. A vector field-based guidance law is employed to drive each UAV toward the reference path. A rotational repulsion mechanism, utilizing relative distance and bearing between UAVs, is proposed to avoid collisions during convergence to the path, and an inter-UAV spacing error-based velocity control law is presented to achieve uniform separation along the path. Analytical guarantees are established for the convergence of the inter-UAV spacing errors to zero, ensuring uniform separation along the path. Numerical simulations demonstrate the efficacy of the proposed method.

Keywords: Robotics, Multi-agent and Networked Systems, Modeling and Validation
Abstract: This paper considers the dynamic modeling of a mechanical system comprising multiple wheeled mobile robots moving atop a shared flat platform with inertial dynamics of its own. Two variants of such a system are discussed: a simpler version in which carts roll along a one-dimensional platform and a more complicated version in which jointed snake-like robots roll atop a two-dimensional platform. In the latter case, nonholonomic constraints couple the robots to the platform. Symmetries associated with conservation laws are exploited for model reduction in both cases; Gibbs--Appell formalism is applied subsequently to obtain reduced-order equations of motion in the latter case. In each case, the robots are coupled to one another indirectly through the platform's movements. We examine control problems in which robots with internal actuation manipulate the dynamics of purely passive robots by exploiting their simultaneous coupling to the platform. We also consider the emergent behavior of systems of passive robots in which internal actuation is replaced by elasticity, and the platform becomes a medium for the exchange of elastic and kinetic energy. The influence of coupling upon the qualitative behavior of individual robots can be made more or less pronounced by varying system parameters; we analyze this influence in terms of energy flow among system components and among basic modes of behavior. We note in particular that when two spring-loaded snake-like robots are released from rest on a common platform, passive mobility of the platform increases the robots' tendency to align their directions of translation over time.

Keywords: Human-Machine and Human-Robot Systems, Robotics, Modeling and Validation
Abstract: Hand sign recognition systems can facilitate intuitive visual communication between humans and robots. Current recognition models often lack validation in outdoor settings, and thus fall short toward field deployment on mobile platforms. In this work, we assess the precision and recall of hand sign recognition models in the field considering robot movement, distance and viewing angle of the human from the mobile vision system. Supervised models were custom trained on skeletal hand data and their F1 scores were compared with various available pre-trained models at varying distances and viewing angles. Statistical analysis presented in this paper shows that the distance to subject from the vision system had a statistically significant impact on hand sign recognition in outdoor environments with mobile robots, while the impact of viewing angle remained insignificant with the models tested.

Keywords: Automotive Systems, Modeling and Validation, Estimation
Abstract: This paper presents a vehicle dynamics modelling and tire forces estimation approach utilizing 10 degrees-of-freedom vehicle model, combined slip tire model, and nominal kinematic & compliance (K&C) data-based steering model. The vehicle model includes vehicle’s linear velocities in all 3-axes – longitudinal, lateral & vertical, rotational rates around the 3-axes – roll, pitch & yaw, and four wheel rotational velocities. The combined slip tire model uses Magic formula as nonlinear functions of normal load, slip ratio, slip angle, and camber. The slip ratio is calculated using cumulative torque and normal force acting at each wheel measured by external wheel force transducer system (WFTS) instrumented at each wheel. The measured wheel torques are also used as input to the system. The system model is implemented in unscented Kalman filter (UKF) framework and the estimations are assessed with a vehicle test.

Keywords: Automotive Systems, Modeling and Validation
Abstract: A tire blowout rapidly destabilizes vehicle lateral and yaw dynamics due to a sudden loss of tire force capability. Although different tire parameters and blowout models were studied, the influence of camber angles and thrusts during a tire blowout remains unclear. Through modeling common preset camber angles, this study develops a new tire blowout model to describe vehicle motions after tire blowouts. Simulation results indicate that preset camber angles and camber thrusts will introduce asymmetric lateral forces and yaw moments after tire blowout. For front tire blowouts, the camber-induced imbalance amplifies the lateral deviation toward the blowout side. For rear tire blowouts, the vehicle deviation direction can be reversed due to force and moment variations induced by camber angles and vehicle states. These findings suggest that camber angles and camber thrusts should be considered to enhance the accuracy of tire blowout modeling.

Keywords: Modeling and Validation, Estimation, Mechatronic Systems
Abstract: Crane tip-over accidents pose a significant risk, often resulting in catastrophic injuries and considerable financial losses. To better understand these accidents and develop mitigation technology, a reliable model validated with data from actual accidents is needed. Video tracking software can be used to extract crane motions from bystander videos. This paper compares data from a software-based video tracking method to measurements from tip-overs of a small-scale 3D-printed mobile crane. Tests with an Inertial Measurement Unit (IMU) showed strong correlation between video-derived boom angles and direct IMU measurements. The results were also compared against a SolidWorks Motion Analysis Model.

Keywords: Modeling and Validation, Mechatronic Systems, Motion and Vibration Control
Abstract: Turntable ladders are essential vehicles for fire departments worldwide. Due to lightweight construction, mechanical oscillations occur in the ladder. Over the years, model-based methods for active oscillation damping have proven to be effective and valuable assistance functions. To reduce the measurement effort for parameter identification, a Digital Model is proposed that allows for a simulation-based parametrization of the control system. Since the hydraulic actuator dynamics significantly affect control performance, a precise dynamic model of the hydraulic system is developed and validated using experimental data. The proposed model of the hydraulics captures asymmetric dead-time delays of the cylinder's motion for both directions through detailed modeling of the load-control valve. By combining a flexible multibody system for the mechanics, the proposed hydraulic model and a model-based approach to Oscillation Damping Control, a Digital Model enables efficient simulation-based controller gain parametrization. The resulting damping in simulation is validated by experimental measurements.

Keywords: Robotics, Modeling and Validation
Abstract: Legged robots have demonstrated promising agility and mobility in unstructured environments. However, highly yielding terrain such as sand and mud brings significant uncertainties and challenges to bipedal locomotion and navigation. Simulation of yielding terrain requires a great demand for accessible sources of computation- and training-friendly contact solvers of yielding terrain modeling. We present a versatile contact simulation tool that features physics-based terramechanics knowledge of two representative yielding substrates, including dry granular and wet mud. The proposed simulation tool captures intrinsic characteristics of yielding terrain, namely, impact-induced shear jamming of granular media, and time-dependent thixotropy and suction of mud. It offers accurate foot-terrain interactions in terms of spatial ground reaction forces. Comparisons between simulations and hardware experiments demonstrate the promising capability to generate diverse and realistic yielding terrain scenarios of the proposed simulator.

Keywords: Control Applications, Estimation, Intelligent Autonomous Vehicles
Abstract: Visual SLAM systems deployed on resource-constrained edge hardware exhibit stochastic processing latency that is coupled to vehicle motion: high-curvature manoeuvres trigger Local Bundle Adjustment, producing pose update delays of up to 600 ms. Standard closed-loop controllers that apply a Zero-Order Hold during these blackouts accumulate sufficient cross-track error to cause trajectory divergence. This paper presents a Dual-Layer Delay-Compensation framework comprising a Latency-Aware Speed Scheduling (LASS) law and an IMU dead-reckoning state predictor, coupled to a receding-horizon Model Predictive Controller. Hardware profiling of ORB-SLAM3 on the custom 1/10th-scale electric vehicle (DEFT) platform and Hardware-in-the-Loop simulations over three trajectory geometries confirm that the proposed framework eliminates crashes and maintains post-spike mean cross-track error below 0.10 m, compared to divergence within 900 ms - 1240 ms for an open-loop baseline maneuver.

Keywords: Control Applications, Control Design, Optimal Control
Abstract: Current ripple is known to reduce the efficiency of electric drive systems, as it contributes to inverter losses, noise, vibration, and electromagnetic interference. In this work, we propose a hierarchical control structure combining a model predictive current regulator with a ripple-aware modulation scheme that minimizes the current ripple at each sampling period. Compared with the standard proportional-integral current regulator cascaded with the conventional space vector modulation, the proposed approach achieves a significant reduction in both current ripple and switching transitions while maintaining low computational complexity. Simulation tests across multiple operating points demonstrate the method's effectiveness.

Keywords: Control Design, Control Applications, Optimal Control
Abstract: This paper presents a method to generate PID controllers using a state-space formulation. It shows that a continuous-time (CT) PID can be generated for any first or second-order, CT, linear, time-invariant (LTI) model. The work starts with a comparison of the standard PID form and a full-state feedback regulator. The P, I, and D terms can be arranged as states of an augmented second-order system. From a single measurement of the plant output, the state and state derivative can be generated from a continuous-time estimator. Finally, we will show how to add the reference into this state-space structure to complete the PID, but with improved filtering. A large advantage of this method over conventional PID is that the derivative term is generated naturally through the state matrix, rather than a raw differentiator, promising improved noise rejection. The difficulties of formulating some of the more general forms of this structure give a lot of insight into both this method and conventional PID controllers themselves. An illustrative design example illustrates the performance advantages of this approach.

Keywords: Aerospace, Control Applications
Abstract: This paper proposes a deep reinforcement learning (RL) based solution for the target enclosing problem for a non-holonomic unmanned aerial vehicle (UAV) operating under limited sensing and system lag. Our approach enables an end-to-end learning-based agent to autonomously interact with the environment and develop enclosing strategies that ensure safety and containment around a stationary target. To better account for temporal dependencies arising from vehicle dynamics and autopilot lag, we propose a history-augmented policy architecture based on Proximal Policy Optimization (HA-PPO) that incorporates observation history into the policy input. A carefully designed reward function guides the learning agent by combining three reward components: a quadratic distance-based reward penalizing deviation from the desired enclosing radius, a state-dependent velocity reward that promotes stability during radial transitions, and an acceleration penalty to enforce a smooth trajectory. Thus, the agent learns to trade off between maintaining the desired enclosing geometry and mitigating aggressive control responses, especially in the presence of autopilot lag. Our results demonstrate that the learned policy, achieved through our carefully designed reward function, achieves reliable target enclosing performance. Moreover, incorporating observation history leads to improved stability and better generalization ability as compared to standard PPO.

Keywords: Machine Learning in modeling, estimation, and control, Estimation, Stochastic Systems
Abstract: Latent-space manipulation in generative models typically changes multiple output properties at once because the decoder entangles them. This paper introduces a Jacobian-based controller that steers individual decoded graph properties while suppressing off-target drift. Starting from a class-conditioned graph variational autoencoder, we define four decoded spatial scores and compute their local Jacobian with respect to the latent code. A regularized least-squares update moves the target score toward a specified value while holding the remaining scores near their initial values. Across validation-selected checkpoints from three independently trained models per domain, all five sensor classes, and all ten MNIST digits, the controller gives median selectivity gains of 43–96× on sensor layouts and 36–292× on MNIST for coverage, signal uniformity, and spatial balance, with additional gains over a derivative-free CMA-ES latent optimizer. The connectivity proxy is a mixed case: selectivity improves on sensor layouts but remains below one on MNIST, even though target error and off-target drift reduce by one to three orders of magnitude. A singular-value analysis of the score Jacobian explains why the controller helps on some attributes more than others and provides a diagnostic for controllability of the learned decoder.

Keywords: Robotics, Control Applications, Human-Machine and Human-Robot Systems
Abstract: This paper presents the design and development of the fourth-generation Daily Assistive Stroke Patient Exoskeleton (DASPE), an upper-limb assistive device for real-time motion tracking. The system integrates a cable-driven mechanical design with a predictive incremental PID controller to reduce tracking lag. Experimental evaluation across varying speeds and tip loads shows low tracking error at speeds up to 10 deg/s. Performance degrades at higher speeds due to control latency, while results indicate minimal sensitivity to external loading. These findings highlight the need for adaptive control strategies to further improve tracking performance.

Keywords: Control Applications, Robotics, Manufacturing Systems
Abstract: This paper compares impedance control and model predictive impedance control (MPIC) for assembling interlocking metasurfaces (ILMS) using a six degree-of-freedom robotic arm. Environment system parameters used in the MPIC were identified by experimentally modeling ILMS as a mass-spring-damper using position and force/torque data. Results show that both the impedance controller and MPIC were effective strategies to assemble the ILMS. The MPIC reduced assembly force in the z-direction by 89.3% compared to position control and 56.3% compared to impedance control. We evaluated controller stability with passivity analysis and found both impedance control and MPIC to be stable.

Keywords: Mechatronic Systems, Machine Learning in modeling, estimation, and control, Manufacturing Systems
Abstract: Wire Arc Additive Manufacturing is a layer-wise metal additive manufacturing process that uses an electrical arc to melt wire feedstock. A range of process planning and control approaches have been proposed for WAAM to improve geometric accuracy using static maps between process inputs and deposited bead geometry, which do not capture process dynamics. Here, we propose an in-layer predictive controller based on an innovation-driven estimator. The estimator uses a long short-term memory (LSTM) recurrent neural network to model the process dynamics, predicting the deposition height as a function of the previous state, torch speed input, and feedback sensor measurement of the height. The innovation-driven LSTM shows improved prediction performance over a conventional LSTM without innovation feedback. For control, the LSTM is linearized such that it can be used to solve a quadratic program to determine the next control input. The in-layer controller is designed to track a reference deposition profile determined between layers. The performance of the controller is demonstrated experimentally on a robotic WAAM testbed, showing 47% improvement in height error over a layer-wise control approach using a static map.

Keywords: Modeling and Validation, Manufacturing Systems, Mining, Mineral and Metal Processing
Abstract: Thin webs in roll-to-roll (R2R) manufacturing often exhibit out-of-plane waviness caused by cross-direction variations in developed machine-direction length. This paper presents a geometry-based, bidirectional model that relates widthwise length distributions to the resulting three-dimensional edge wavy shapes between rollers. Two complementary formulations are introduced and developed: A spherical construction using a thin-shell mid-surface representation and a cylindrical construction to determine boundary conditions for U-wrap and S-wrap configurations. The model predicts admissible waviness and inversely identifies compatible length profiles from measured surfaces. The framework provides a constructive, load-independent description of edge waviness without requiring material constitutive assumptions.

Keywords: Optimal Control, Adaptive and Learning Systems, Manufacturing Systems
Abstract: Injection Molding (IM) is a high-volume production process for plastic parts, in which in-mold cavity pressure strongly influences final part quality. While prior work has successfully developed a spline-based Norm-Optimal Iterative Learning Control (NOILC) approach for cavity-pressure control, manually tuning NOILC parameters remains challenging and heavily relies on control engineering expertise. We propose an automated, batch-learning Bayesian Optimization (BO) method to tune the controller parameters of NOILC. The approach optimizes controller parameters to minimize cavity-pressure tracking error and cavity-pressure overshoots across IM cycles. We validate the approach using a simulation model, specifically altering material viscosity to emulate the batch-to-batch fluctuations characteristic of processing Post-Consumer Recycled (PCR) materials. In a first proof of concept, the BO framework successfully identifies an improved controller parameterization within only 15 BO iterations in simulation. The optimized controller exhibits improved trajectory tracking and limits pressure overshoots to below 0.3 bar during simulated PCR processing. By using non-equidistant nodes in the spline-based NOILC, we achieve a 74.75 % reduction in the number of decision variables, thereby heavily reducing the computational complexity compared to the baseline NOILC parameterization.

Keywords: Optimal Control, Robotics, Machine Learning in modeling, estimation, and control
Abstract: Nonholonomic mobile robots are increasingly used for autonomous transportation and material handling in manufacturing plants and warehouses. These systems range from single-body robots, commonly represented by unicycle models, to articulated vehicles that pull multiple trailers. Their nonlinear dynamics, actuator limitations, and coupled motion make accurate, collision-free control in real time particularly challenging. Nonlinear model predictive control (NMPC) is a standard approach to solving nonlinear optimal control problems, but its online computational cost can be prohibitive because it requires repeated nonlinear optimization. We present a learning-based framework for real-time constrained optimal control that shifts computation offline. The approach learns a neural co-state (adjoint) policy that encodes optimality through the system Hamiltonian rather than directly approximating a control law. Feasibility is enforced separately at runtime through an efficient convex optimization that incorporates actuator limits and safety constraints expressed as control barrier functions. The resulting controller satisfies these constraints while retaining the optimality structure encoded by the learned adjoint. On constrained control tasks involving a unicycle model, our framework achieves performance comparable to NMPC at more than two orders of magnitude lower computational cost. It also significantly outperforms reinforcement learning methods in out-of-training-distribution scenarios. To investigate its applicability to higher-dimensional nonholonomic systems, we extend the framework to tractor–(n)-trailer systems. Our simulation studies evaluate goal regulation from both seen and out-of-distribution initial conditions, with and without obstacles. This extension demonstrates the potential of adjoint-based learning for computationally efficient, safety-critical control of articulated autonomous vehicles used in advanced manufacturing.

Keywords: Robotics, Mechatronic Systems, Uncertain Systems and Robust Control
Abstract: Simulation to reality (sim2real) transfer from a dynamics and controls perspective usually involves re-tuning or adapting the designed algorithms to suit real-world operating conditions, which often violates the performance guarantees established originally. This work presents a generalizable framework for achieving reliable sim2real transfer of autonomy-oriented control systems using multimodel multiobjective robust optimal control synthesis, which lends well to uncertainty handling and disturbance rejection with theoretical guarantees. Particularly, this work is centered around a novel actuation-redundant scaled autonomous vehicle called Nigel, with independent all-wheel drive and independent all-wheel steering architecture, whose enhanced configuration space bodes well for robust control applications. To this end, we present the mechatronic design, dynamics modeling, parameter identification, and robust stabilizing as well as tracking control of Nigel using the proposed framework, with exhaustive experimentation and benchmarking in simulation as well as real-world settings.

Keywords: Robotics, Optimal Control, Modeling and Validation
Abstract: Trajectory optimization libraries (TOLs) for robotics and related fields are predominantly implemented using object-oriented programming (OOP) paradigms, yet the implications of this design choice for code size, reproducibility, and performance have not been systematically studied. In this paper, we conduct a case study comparing a representative OOP-based TOL to our implementation in a declarative algebraic modeling language (AML). The aim is to explore how paradigms affect TOL implementations when comparing the dominant OOP paradigm in robotics with a paradigm widely used in the broader optimization community.

Keywords: Robotics, Modeling and Validation, Motion and Vibration Control
Abstract: This paper presents a simulation model of UR10e robots including link and joint deflections, and illustrates how the model can be used for improving the path accuracy and accuracy during interaction with an environment. The deflection parameters are estimated from experiments and shown to work across robots of the same type. Link and joint deflections are particularly imminent for collaborative robots, where joints and in particular links are less stiff. This calls for better kinematic models that include varying deflections due to varying forces and torques on links and joints. In this paper, the focus is on deflections of small selected set of most significant deflections, and it is shown that both joint and link deflections can be very accurately modelled and that the deflection models are directly transferable to other UR10e robots. As an example of use, we show that the mean path accuracy for compensating an external load of 8 kg can be improved by more than a factor 4.

Keywords: Robotics, Modeling and Validation, Mechatronic Systems
Abstract: This work presents an adaptive 3D scanning framework combining robotic manipulation with dual-rotation turntable reconstruction for complete surface coverage. Conventional scanners cannot capture bottom surfaces due to platform contact and lack real-time feedback. The system integrates a Kinova Gen3 manipulator with a novel dual-rotation turntable featuring independent camera and platform motors. The robot performs rapid scanning for immediate feedback, autonomously flips objects to expose occluded surfaces, and then coordinates with the turntable for high-fidelity reconstruction. Simulation demonstrates 95% coverage versus 65% conventional, while physical prototype validation confirms mechanical operation and ROS integration. The framework supports sequential or simultaneous modes for manufacturing quality control applications.

Keywords: Robotics, Underwater Vehicles, Estimation
Abstract: This paper introduces a relative localization method that leverages the directivity of Sub-GHz radio antennas and fuses data from an inertial measurement unit to enhance position estimation for a swarm of robots performing formation control. In contrast to purely inertial approaches, the proposed framework mitigates the inherent long-term drift while remaining low-cost, compact, and lightweight compared with a USBL system. The localization device comprises four flat, patch-type RF antennas and a motion sensor forming a 9-axis IMU. The paper also addresses the design challenges associated with enabling RF operation underwater and realizing a low-frequency directional antenna without sacrificing compactness. The resulting device measures 15 cm times 15 cm times 8 cm. An experimental setup is developed to acquire sensor data together with ground-truth measurements obtained via a computer vision system. Physics-aware training applied to the collected data demonstrates that the proposed framework can achieve a mean absolute localization error of 0.24 m over a 3 m range and 0.64 m over a 6 m range.

Keywords: Soft Robotics, Estimation, Sensors and Actuators
Abstract: Physical reservoir computing (PRC) refers to the use of a physical dynamical system as a computational resource for tasks such as state estimation and control, but there has been a lack of formal study of design rules towards more effective design of such physical reservoirs. Using a pneumatic soft arm with a five-pouch sensing column, this work studies how the pouch interconnection topology, robot stiffness, and the number of instrumented sensors affect bending-angle estimation performance. Across 36 matched trials spanning waveform, baseline pressure of the sensing column, and actuation range, all designs are evaluated under the same-time bending-angle estimation benchmark using 0.2~s of pressure history and a fixed ridge estimator. Our analysis of the experimental results leads to three design guidelines. First, independently sealed pouches preserve a much richer observable state than a shared manifold. Second, increasing the baseline pressure of the sensing column makes the pouch responses more redundant and increases estimation error most strongly in the coupled topology. Third, in the sealed topology, two strategically placed sensors already recover most of the attainable benefit, three capture essentially all of it, and additional sensors provide little or no additional value. In summary, the results suggest that topology, stiffness, and number of instrumented sensors should be co-designed for accurate PRC of soft robot states; stronger excitation alone cannot recover the diversity that poor design choices have already removed.

Keywords: Soft Robotics, Sensors and Actuators, Modeling and Validation
Abstract: Twisted string actuators (TSAs) provide unique benefits over many existing compliant actuators. However, they are limited to around 30% maximum strain output. The recently created twisted and coiled string actuators (TCSAs) demonstrated approximately 70% strain. Evolved from TSAs, TCSAs produce actuation by first twisting a pair of strings using a motor and then coiling the fully twisted strings. Considering that TCSAs largely retain TSAs’ advantageous properties, it is highly promising to obtain lightweight and compact soft robots driven by TCSAs. For the first time, this study successfully demonstrates a TCSA-driven soft robotic manipulator. Using a separator-design for the TCSA, the soft manipulator produces smooth bending motion while being compact. A kinematic model is proposed to capture the relationship between the manipulator’s bending angle and the TCSA’s motor rotation angle. Preliminary experimental validation shows promising results, for example, the root mean square error (RMSE) of the model was 1.70◦, while the range of the bending angle was 90◦.

Keywords: Education for modeling, estimation and control, Chemical Process Control, Distributed Parameter Systems
Abstract: Pipelines are essential components of industrial infrastructure, enabling the transport of water, energy, and other critical resources. Ensuring their long-term operational integrity is vital for sustainability, minimizing resource losses and preventing environmental harm. This experimental realization presents a Kalman Filter–based approach for leak detection and localization in a gradually inclined pipeline manifold. The system is spatially discretized into multiple segments, each modeled using continuity and momentum conservation equations. The Extended Kalman filter (EKF) framework then estimates flow rates, pressures, and potential leak parameters, including location and leak rate. The method provides good estimates of the leak location and leak rate simultaneously.

Keywords: Robotics, Assistive and Rehabilitation Robotics, Modelling and Control of Biomedical Systems
Abstract: The role of Golgi tendon organs (GTOs) in force feedback during voluntary movement remains difficult to isolate experimentally. This study provides computational and experimental evidence that interaction-force regulation under variable environmental stiffness is consistent with a force-feedback mechanism potentially involving GTO pathways. First, a neuromusculoskeletal model of the monoarticular elbow flexor muscle group was trained to achieve target forces of 40, 50, and 60 N under fixed stiffness and tested at untrained stiffness levels using open- and closed-loop configurations. Closed-loop models reduced steady-state force deviations by 10.4%, 12.1%, and 13.0% for M40, M50, and M60, respectively, compared with open-loop models; open-loop performance was significantly affected by stiffness. Second, human experiments used an instrumented elbow exoskeleton under three stiffness levels and three trial types (reference, blind, and catch), with stiffness randomized during catch trials. Results showed consistent force regulation, with no significant effects of stiffness or trial type. Agreement between simulations and experiments is consistent with force-feedback regulation but does not isolate GTO contributions from other somatosensory inputs. These findings support a functional role for force feedback in stabilizing interaction forces and inform neuroprosthetic and rehabilitation design.

Keywords: Linear Control Systems, Control Applications, Modelling, Identification and Signal Processing
Abstract: In this paper, we provide counterexamples to the statement in the current literature matrix theory and control theory textbooks that eigenvalues are invariant under arbitrary, linear non-singular transformations and such transformations are called similarity transformations. In this paper, we argue that not all linear, non-singular transformations are required to be Similarity transformations and that there can be user-designed non-similarity transformations under which eigenvalues become Variant under those transformations. In addition, we propose a single transformation matrix, which belongs neither to the similarity category nor to the non-similarity category. In other words, by careful user-selected A and Z matrices, in the relationship Az = Z AZ −1 (or Az = Z −1 AZ ) where Az is the transformed matrix, we demonstrate that the same transformation matrix Z shows Azleft = Z−1AZ as having same eigenvalues as A but the other matrix Azright = Z∗A∗Z−1 has different eigenvalues than of A. It is possible that the nature of Azleft and Azright are interchangeable. Such One Sided Eigenvalue Invariance/Variance can be analytically explained as the result of Commutativity Constraint satisfaction/violation. Hence, finally, it is hoped that the counterexamples in this paper will unequivocally put to rest the argument that such eigenvalue variance (we have been routinely observing but ignored) is due to numerical sensitivity issues. It is not, it is rather due to Commutativity Constraint violation. It is well known that in matrix theory, product of two matrices (even when both are real square matrices), A∗B is not necessarily equal to the product B ∗ A.

Keywords: Modeling and Validation, Distributed Parameter Systems
Abstract: This paper presents a macroscopic approach for modeling pedestrian movement. A convection-diffusion partial differential equation (PDE) is described and used to model the flow of runners along a marathon course. A marathon is a good modeling case study for human movement because the location of runners is only known in sparse locations, so an accurate model is needed to know where people are at all times. As well, marathons offer many safety and logistical problems that can be solved by modeling human movement. Data is obtained from the 2023 Columbus Marathon, and model improvements are made to include a temporally varying convection term. Results show an increase in model accuracy which can allow for better resource allocation and safety measures.

Keywords: Modelling, Identification and Signal Processing, Estimation, Stochastic Systems
Abstract: This paper presents a model-based statistical framework for detecting catalyst aging in diesel engine SCR systems. A dynamic model for NOx reduction under catalyst saturation with ammonia is derived from molar conservation, accounting for the interplay between residence time and sensor sampling frequency. The model parameters are estimated via maximum likelihood using the bounding relationship between actual and predicted NOx reduction under saturation while handling uncertainties from NOx sensor cross-sensitivity in the absence of ammonia sensors. The estimated parameters are shown to be sensitive to the catalyst's aging state, enabling a Wald test based hypothesis-testing framework for aging detection. The proposed detector is validated using test-cell data at different aging levels and real-world data from four long-haul trucks.

Keywords: Path Planning and Motion Control, Optimal Control
Abstract: This paper investigates the problem of computing the shortest path between two states under resource constraints in environments with resource-replenishment regions. Namely, the length of the path is limited by a budget that can be restored within polytopic replenishment regions. We show that the optimal path in this problem exhibits a distinct geometric structure: it consists of straight-line segments, changes direction at the edges of replenishment regions, and enters a sequence of replenishment regions in a feasible-connected space. We propose an approach to solve the continuous problem in two steps: using a graph-based approach, followed by convex programming. First, we define a graph whose nodes are possible waypoints of feasible paths, and the edges are the Euclidean distances between these nodes. To obtain a discrete set of nodes that ensure a feasible and near-optimal solution, we utilize a wavefront algorithm. With a sufficiently small spacing between wavefronts, the solution of the shortest path problem on this graph yields the optimal sequence of polytopes to visit. Next, we use convex optimization on this sequence of polytopes to refine the solution to optimality. A numerical experiment is presented to demonstrate the effectiveness of the approach.