Keywords: Control applications, Energy Systems, Markov processes
Abstract: Residential islanded microgrids integrating renewable energy and storage can improve resilience, but their operation is challenged by the randomness of solar generation and the need to protect battery health. In this study, we first introduce a multi-battery architecture with vehicle-to-home (V2H) capability, enabling coordinated use of stationary and mobile storage. Second, we evaluate Jump Linear Quadratic Control (JLQC) for energy management of this system, where solar energy randomness is modeled as a Markov chain. Additionally, JLQC's performance is compared with Model Predictive Control (MPC) and Linear Quadratic Regulator (LQR). Simulations of a single-home microgrid assess supply–demand balance, constraint adherence, and battery state-of-charge behavior. Results show that JLQC achieves comparable or improved accuracy relative to MPC while maintaining online operational simplicity similar to LQR, highlighting its effectiveness for practical, real-time energy management in residential microgrids.

Keywords: Identification, Modeling, Reduced order modeling
Abstract: Sparse Identification of Nonlinear Dynamics (SINDy) is a powerful method for discovering parsimonious governing equations from data, but it often requires expert tuning of candidate libraries. We propose an LLM-aided SINDy pipeline that iteratively refines candidate equations using a large language model (LLM) in the loop through in-context learning. The pipeline begins with a baseline SINDy model fit using an adaptive library and then enters a LLM-guided refinement cycle. At each iteration, the current best equations, error metrics, and domain-specific constraints are summarized in a prompt to the LLM, which suggests new equation structures. These candidate equations are parsed against a defined symbolic form and evaluated on training and test data. The pipeline uses simulation-based error as a primary metric, but also assesses structural similarity to ground truth, including matching functional forms, key terms, couplings, qualitative behavior. An iterative stopping criterion ends refinement early if test error falls below a threshold (NRMSE < 0.1) or if a maximum of 10 iterations is reached. Finally, the best model is selected, and we evaluate this LLM-aided SINDy on 63 dynamical system datasets (ODEBench) and march leuba model for boiling nuclear reactor. The results are compared against classical SINDy and show the LLM-loop consistently improves symbolic recovery with higher equation similarity to ground truth and lower test RMSE than baseline SINDy for cases with complex dynamics. This work demonstrates that an LLM can effectively guide SINDy’s search through equation space, integrating data-driven error feedback with domain-inspired symbolic reasoning to discover governing equations that are not only accurate but also structurally interpretable.

Keywords: Manufacturing systems, Sensors, Distributed parameter systems
Abstract: The mechanical properties resulting from additive manufacturing of metal are significantly influenced by the temperature history during the manufacturing process. Consequently, the measurement and control of the temperature have primary importance to ensure part quality. Although the frequency content of temperature signals can dictate both sensor frame rate selection and controller design, time-frequency analysis has received little attention in metal additive manufacturing. This work presents a preliminary study of two direct-laser-deposition (DED-LB/M) experiments in which the laser heat flux follows a Gaussian intensity distribution. Using continuous wavelet transform scalograms, the laser spot size and traverse speed are related to the bandwidth of the measured surface temperature. The derived empirical formulas are specific to the Gaussian-laser DED-LB/M set-up studied, but the methodology can be transferred to other metal-AM processes.

Keywords: Marine/underwater robotics, Predictive control, Identification
Abstract: Energy-efficient trajectory tracking for surface vessels commonly relies on propulsion power models that neglect inflow, and therefore changes in the propeller operating point. For vessels operating near a single nominal condition this is a reasonable simplification, but for overactuated inland vessels with azimuth thrusters that turn, stop, and accelerate frequently, the operating point shifts continually and the unmodeled power variations translate into avoidable energy use. On electric vessels, this dependence is inexpensive to recover: thruster power is directly measurable at the inverter DC side, without dedicated instrumentation. We exploit this to experimentally identify an inflow-aware thruster power model and use it as the economic stage cost of a nonlinear model predictive control (NMPC) trajectory tracking controller with embedded thrust allocation and actuator constraints. We evaluate the approach on an electric inland test vessel, benchmarking it against an otherwise identical NMPC controller with a conventional cubic shaft-speed model as its stage cost. In simulation, the inflow-aware formulation yields only a small benefit of 2.13 percent in near-nominal sailing but reduces energy consumption by 11.3 percent on a maneuver-intensive ferry route. Further field experiments show an average energy reduction of 13.4 percent, consistent with the simulated result.

Keywords: Power Electronics, Automotive applications, Modeling
Abstract: This paper proposes a control-oriented modeling and design framework for the matrix-type dual active bridge (MT-DAB), a single-stage isolated ac–dc converter. The input matrix bridge is modeled and controlled as a space-vector–controlled current-source inverter, while the secondary stage is treated as a dual-phase-shift (DPS) controlled DAB. A generalized averaged large-signal model is derived by exploiting the close correspondence between the MT-DAB’s selected pulse pattern and conventional DPS modulation. Linearization around operating points yields a compact linear time-invariant (LTI) small-signal model that captures the dominant dynamics with only the first harmonic. The models enable systematic synthesis of control loops for independent regulation of output power and input power factor, eliminating the need for precalculated switching instants. Validation against switched PLECS simulations shows high fidelity: the inductor-current dynamics are reproduced with near-exact agreement, and output-voltage transients are predicted accurately for moderate steps.

Keywords: Linear robust control, LMIs, Linear parameter-varying systems
Abstract: We propose a zonotopic framework for synthesizing a single robust state feedback controller that is certified to stabilize every plant inside a matrix zonotope, describing linearly varying parameters or parametric uncertainty. Common robust design strategies rely on checking many vertex models or on complex gain-scheduling, leading to high offline computation and implementation complexity. Our approach finds a single gain that is provably valid across the entire parameter domain, which is simpler to implement and can reduce conservatism by exploiting the structure of the zonotope. We formulate the robust synthesis as a single convex program tailored to the zonotope representation and incorporate practical performance requirements (actuator constraints, decay rate, disturbance attenuation) into the same synthesis stage. In numerical experiments on a representative 4-state example, our controller provides larger stability coverage across the parameter domain, attains comparable transient performance and control effort to more complex designs, and significantly reduces the number and scale of offline synthesis problems required by other robust approaches, compared to common-vertex gain, (H_{infty}), and (mu)-synthesis baselines.

Keywords: Nonlinear systems, Optimization, Control applications
Abstract: Recently, the frequency and severity of heavy-rainfall disasters have increased, making the reduction of precipitation through weather control technology an urgent priority. This study proposes a sequential optimization method that accounts for model nonlinearity and a hierarchical sparse optimization method to localize the atmospheric state perturbation for precipitation control in nonlinear weather models. Localization of perturbation regions is promoted through sparse optimization using l1-norm minimization. In addition, a hierarchical optimization algorithm is developed to reduce the size of the regions where perturbations are applied. Validation through a warm bubble experiment using SCALE-RM, a nonlinear numerical weather prediction model, demonstrated that the proposed method reduced the peak precipitation to 40% of its nominal value under conditions where model nonlinearity cannot be ignored. Moreover, the resulting optimal perturbation was confined to highly localized regions.

Keywords: Nonlinear systems, Predictive control, PID control
Abstract: For linear plants, setpoint (step) command following with zero asymptotic error can be achieved with integral control. Within the context of model predictive control without an integrator in the loop, asymptotic setpoint command following requires exact knowledge of the DC gain of the plant. This paper revisits these properties for Wiener plants, which, due to the output nonlinearity, do not have a DC gain. It is shown that the equivalent DC gain of Wiener systems is the command-dependent DC gain. Numerical experiments show that without a prior model, predictive cost adaptive control using linear system identification achieves asymptotic setpoint command following by learning the command-dependent DC gain. The approach is illustrated by a two-degree-of-freedom structure with measurements given by a capacitive position sensor.

Keywords: Numerical analysis, Linear robust control, Optimization
Abstract: Gyroscopic interconnections enable redistribution of energy among degrees of freedom while preserving passivity and total energy, and they play a central role in controlled Lagrangian methods and IDA–PBC. Yet their quantitative effect on transient energy exchange and subsystem performance is not well characterised. We study a conservative mechanical system with constant skew‑symmetric velocity coupling. Its dynamics are integrable and evolve on invariant two‑tori, whose projections onto subsystem phase planes provide a geometric description of energy exchange. When the ratio of normal‑mode frequencies is rational, these projections become closed resonant Lissajous curves, enabling structured analysis of subsystem trajectories. To quantify subsystem behaviour, we introduce the inscribed‑radius metric: the radius of the largest origin‑centred circle contained in a projected trajectory. This gives a lower bound on attainable subsystem energy and acts as an internal performance measure. We derive resonance conditions and develop an efficient method to compute or certify the inscribed radius without time‑domain simulation. Our results show that low‑order resonances can strongly restrict energy depletion through phase‑locking, whereas high‑order resonances recover conservative bounds. These insights lead to an explicit interconnection‑shaping design framework for both energy absorption and containment control strategies, while taking responsiveness into account.

Keywords: Time delays, PID control, Control applications
Abstract: We present a tuning algorithm for linear time-invariant (LTI) control loops with arbitrary transport delay. The method adjusts the closed-loop gain of an existing baseline controller to achieve a specified damping of the dominant plant dynamics while preserving stability and the exact transport delay throughout the tuning process. The approach is based on root locus principles but replaces explicit root computation with a significantly leaner, one-dimensional root-finding problem. The effectiveness of the proposed method and the advantages of retaining the exact delay over delay approximations are demonstrated on two real-world plants.

Keywords: Power systems, Machine learning, Cyberphysical systems
Abstract: Critical infrastructures are becoming increasingly complex as our society becomes increasingly dependent on them. This complexity opens the door to new possibilities for attacks and a need for new defense strategies. Our work focuses on feedback instability attacks on the power grid, wherein an attacker causes cascading outages by introducing unstable dynamics into the system. When stress is placed on the power grid, a standard mitigation approach is load-shedding: the system operator chooses a set of loads to shut off until the situation is resolved. While this technique is standard, there is no systematic approach to choosing which loads will stop an instability attack, and existing approaches can even exacerbate the attack. This paper explores an approach for a data-driven methodology for load shedding decisions. We show a proof of concept on the IEEE 14 Bus System using the Achilles Heel Technologies Power Grid Analyzer, and identify modified Prony analysis (MPA) as a viable method for detecting instability attacks and triggering defense mechanisms. In sum, we provide motivation, proof-of-concept, and architecture for a previously unconsidered mitigation technique for this type of instability attack.

Keywords: Power systems, Optimization, Neural networks
Abstract: Managing multi-building smart grids requires accurate demand forecasting, efficient resource allocation, and robust real-time control under uncertainty. This paper presents an integrated energy management framework that combines deep learning forecasting, metaheuristic optimization, and Model Predictive Control (MPC) for priority-aware energy reallocation. A hybrid CNN-LSTM model captures multivariate temporal dependencies to deliver short-term demand predictions. To maximize accuracy and eliminate manual tuning, a Grey Wolf Optimizer (GWO) fine-tunes the network's hyperparameters and training configurations. These optimized forecasts serve as inputs to a receding-horizon MPC module, which determines cost-effective dispatch decisions that minimize operating costs and peak grid demand while enforcing building-level priority constraints and battery operational limits. Finally, a sequential sensitivity analysis identifies the optimal operational knee point to balance the trade-offs among peak shaving, economic cost, non-critical load tracking, and battery cycling. By coupling optimized learning-based prediction with closed-loop decision-making, the proposed modular and scalable architecture enables resilient microgrid energy management.

Keywords: Renewable Energy, Computer-aided control design
Abstract: Conventional ocean current turbine (OCT) designs follow a sequential approach, optimizing physical parameters before developing control strategies. This separation can lead to suboptimal performance, as the physical design may limit the effectiveness of the control system and vice versa. This paper presents a control co-design approach for a variable-speed dual-rotor OCT, accounting for the plant geometry to optimize geometric parameters using a nested approach within a closed-loop control context. The design space includes generator dimensions (outer diameter, inner diameter, and length) and wing parameters (span, chord, thickness, and pitch angle). Optimization of the generator geometry improves the power-to-weight ratio (PWR) by reducing overall system mass and enhancing buoyancy characteristics. However, wing optimization results in reduced lift and a corresponding degradation in PWR. To compensate for this effect, a finite-horizon optimal controller is employed to regulate the water volumes in the front and rear chambers of the buoyancy tank, restoring lift and adjusting the turbine operating depth and pitch angle, ensuring stable navigation within the water column and continuous operation at the desired depth. By incorporating a control-proxy formulation, the proposed framework captures plant-controller coupling effects without explicitly embedding the controller within the geometric optimization process. Simulation results for a 1200 kW dual-rotor OCT show that this approach enables high-efficiency power extraction with optimized component sizing and active buoyancy regulation.

Keywords: Renewable Energy, Control applications
Abstract: Wind energy is an increasingly growing source of energy. To increase the capacity factor of existing wind farms, and to improve the design of future wind farms, it is important to address the wake effect. Wake interactions negatively impact the efficiency of wind farms, and wind farm flow control techniques are developed to mitigate this. Floating wind farms have the unique possibility of repositioning turbines by controlling the aerodynamic thrust force. This enables the adaptation of the wind farm layout to the wind conditions, maximizing the farm power output. The conventional strategy for turbine repositioning is to reposition turbines to a steady-state optimal solution. While initial results on this technique are promising, it is highlighted that mooring lines need to be lengthened substantially to facilitate sufficient displacement. Recent work has shown that for tauter mooring configurations, the optimal repositioning solutions are dynamic, and involve periodic yaw control signals that result in dynamic turbine repositioning. Dynamic turbine repositioning poses a synchronization problem, as the phase difference between the actuation of the turbines is critical to the effectiveness of the technique. Therefore, this work presents a data-driven solution with extremum seeking control, to control the phase difference for dynamic repositioning, and thereby maximize the wind farm power production. The technique is demonstrated in a mid-fidelity simulation environment, and it is shown that dynamic repositioning attains a higher power output and requires a lower yaw amplitude compared to static repositioning for a wind farm consisting of two 15 MW turbines.

Keywords: Sensors, Energy Storage, Energy Systems
Abstract: Reliable detection of internal short-circuit (ISC) faults in large-format prismatic lithium-ion cells is challenging due to internal parallelization of jellyrolls. This work presents a threshold-based fault detection framework comparing voltage-only, temperature-only, and combined sensing strategies across ISC severities from medium-hard (R_sh/R_0 = 100) to soft (R_sh/R_0 =100,000) under three operating conditions: C/3 constant discharge, rest, and a drive cycle. Thresholds are independently optimized for each condition, with voltage thresholds varying by over two orders of magnitude (3 to 850 mVs) while temperature thresholds remain relatively consistent (1.3 to 2.3 Cs). Voltage-only detection achieves a rapid response during constant discharge and at rest, but degrades severely under dynamic operation, requiring up to 49 minutes for medium-hard shorts and failing entirely for medium and soft shorts at multiple SOC levels. Temperature-only detection maintains consistent performance across all scenarios. Kullback-Leibler divergence analysis confirms that voltage failure for soft shorts reflects fundamental signal indistinguishability. These results demonstrate that temperature sensing is essential for comprehensive fault coverage in parallel-connected cells.

Keywords: Automotive applications, Reinforcement learning, Learning
Abstract: Conventional energy management strategies based on deep reinforcement learning (DRL) often suffer from slow convergence and lack of generalization. This paper proposes an emph{Expert Information Reconstruction-based Policy Learning} (EIRPL) framework to address these issues. Specifically, EIRPL first reconstructs an expert policy from driving data via dynamic programming to train a policy that optimizes energy efficiency, ensuring global optimality for each driving scenario. Then, an implicit action-value function that captures expert knowledge is recovered from the trained optimal policy through adversarial inverse reinforcement learning. Finally, the Soft Actor–Critic (SAC) algorithm is employed to optimize the policy corresponding to the learned value function. Notably, the proposed EIRPL separates the training of the value function and policy into two distinct stages, which accelerates convergence and enhances robustness. Finally, simulation results demonstrate a significant improvement in convergence speed, a higher cumulative reward, and a noticeable reduction in fuel consumption across multiple driving cycles.

Keywords: Autonomous systems, Nonlinear systems, Neural networks
Abstract: This paper studies learning-based synthesis of control barrier functions (CBFs) with an explicit treatment of hard input constraints. Training a task-specific CBF from scratch for each new environment can be computationally expensive and data-inefficient. To address this, we propose a meta-learning approach based on Model-Agnostic Meta-Learning (MAML) that learns an initialization from data collected across a set of related tasks, enabling rapid adaptation to a new environment using only a small amount of task-specific data and a few gradient steps. In closed-loop experiments, a nominal goal-directed control input is computed under input bounds and then filtered by a CBF-QP safety filter to minimally modify the input while enforcing safety under the same actuator limits. Simulation results demonstrate safe, adaptive navigation and good generalization across diverse environments under input constraints.

Keywords: Energy Systems, Reinforcement learning, Predictive control
Abstract: This paper proposes a novel approach to the HVAC problem based on combining MPC with RL in a soft actor-critic scheme. Our algorithm integrates the predictive planning and constraint satisfaction capabilities of MPC with the adaptability of data-driven learning from RL by constructing stochastic policies in which a neural network parameterizes the MPC to output actions within predefined bounds. The resulting algorithm exploits knowledge of future price profiles and weather forecasts while maintaining indoor climate within time-varying bounds to ensure occupant comfort. Given an economic objective, the control algorithm learns to minimize energy cost by adjusting MPC parameters, as demonstrated through a numerical example. The code is open-source and publicly available.

Keywords: Manufacturing systems, Reinforcement learning
Abstract: This paper presents a Deep Reinforcement Learning (DRL)-based control framework for product height regulation in the Directed Energy Deposition (DED) process. A control-oriented, layer-wise height model is first developed to predict incremental layer height, and validated on unseen experimental trials, demonstrating accurate layer height prediction. Building on the identified model, a DRL-based control framework is employed to minimize total deposition time while achieving a specified target product height within tolerance and satisfying actuator and safety constraints. The proposed framework generates layer-wise process parameters, including laser power, traverse velocity, dwell time, and nozzle increment, while also determining the required number of deposition layers. Simulation studies targeting a 15 mm product height with a 0.5 mm tolerance demonstrate robust final height regulation. These results demonstrate that DRL enables accurate and time-efficient height regulation in DED.

Keywords: Nonlinear systems, Nonlinear robust control, Robotics
Abstract: Kernel-based data-enabled predictive control (Kernelized DeePC) enables nonlinear system control directly from data through kernel lifting, but its scalability is limited by the quadratic growth of the Gram matrix with data size. This work proposes a selection-based Kernelized DeePC framework that reduces this burden by optimizing over a dynamically selected subset of informative data using a k-nearest neighbor strategy in the standardized past-window space. The resulting localized formulation preserves the nonlinear modeling capability of kernel methods while significantly reducing computational cost and improving robustness by excluding irrelevant or noisy data. Simulation results on a high-speed Cable-Driven Parallel Robot (CDPR) show that the proposed method achieves accurate and stable tracking under repetitive circular references, outperforming standard data-driven approaches. These results demonstrate that selection-based Kernelized DeePC provides a practical route toward real-time, model-free control of nonlinear fast-sampled systems.

Keywords: Game theory, Reinforcement learning
Abstract: This paper proposes a hybrid framework for crowd evacuation in environments with obstacles by combining the velocity rule obtained in the Hughes model with Reinforcement Learning (RL). We first formulate the evacuation problem as a differential game in the framework of atomic games, and then we present the non-atomic counterpart by assuming that the number of agents tends to infinity. Recent literature shows that the resulting game theory settings retrieve the macroscopic Hughes dynamics. The optimal velocity of agents naturally decomposes into a density-dependent magnitude (speed) and a directional orientation. While the speed is mainly governed by the density, especially within the region surrounding an agent, solving the emerging Hamilton-Jacobi-Bellman (HJB) equation to determine the optimal global potential field, when tackling the constrained problem using dynamic programming, becomes computationally prohibitive in domains with complex obstacles as the feasible region becomes non-convex. To overcome this computational bottleneck, we employ an Actor-Critic RL architecture to approximate the optimal heading for a single representative agent. This approach leverages the Hughes-derived speed regulation and RL for obstacle avoidance. We validate the proposed method via numerical simulations, demonstrating successful evacuation in a complex, non-convex scenario where traditional explicit solutions struggle.

Keywords: Mobile Robots, Neural networks, Localization
Abstract: Autonomous navigation on narrow pedestrian sidewalks presents a significant challenge for mobile robots used in structural health monitoring (SHM) applications. Accurate global and local perception is required to keep the mobile robot within the sidewalk boundaries while avoiding curbs, vegetation, and dynamic obstacles. This paper presents a lower-cost, LiDAR-free embedded robot control system for accurate sidewalk navigation toward predefined global goals without requiring online mapping or prior SLAM runs. Global navigation references were obtained by converting raw OpenStreetMap (OSM) data into a georeferenced waypoint layer, while local path planning is achieved using a custom lightweight semantic segmentation model integrated with the Robot Operating System 2 (ROS 2) Nav2 navigation stack. The proposed approach was first validated in ROS 2-based simulation and subsequently optimized for real-time deployment on a mobile robot platform equipped with low-cost sensing, including a stereo camera, GPS, and IMU, running on an edge computing system. Experimental results demonstrate stable simulation-based navigation, accurate real-time segmentation performance, and the reliability of the OSM-derived coordinates for sidewalk navigation.

Keywords: Mobile Robots, Planning
Abstract: Planar mover systems that use magnetic levitation to move around movers offer great potential to automated assembly lines, laboratories or cleanrooms. A scalable motion planning algorithm that can be used to control different movers operating in the same environment is required to exploit the capabilities of such motion systems. This work proposes a centralized approach that plans trajectories for movers independently, checks for collisions and modifies trajectories if needed and detects and resolves deadlocks. This approach exploits the presence of a central computer with knowledge of all movers and the high tracking accuracy of the system. Through simulation results, we show that the computation time scales well with the number of agents, requiring less than 100ms per iteration for 100 agents in a large environment. Additionally, the method is validated on a real Beckhoff XPlanar mover system.

Keywords: Mobile Robots, Reinforcement learning, Navigation
Abstract: The development of a controller structure relying on Reinforcement Learning (RL) for a mobile inverted spherical pendulum is the subject of this paper. The mobile base has an omnidirectional triangular base, allowing the system to move in an obstacle-cluttered environment. The system is equipped with a LiDAR, an Inertial Management Unit, and an RGB-D camera. The system fuses measurements from these sensors and infers an obstacle-free path based on third-order Bezier curves. The pose of the vehicle uses concepts from Simultaneous Localization And Mapping (SLAM) based on the RTAB-Map algorithm. The RL-controller isolates the tip of the pendulum from the rest of the body and moves the vehicle so as to keep the pendulum in the upright position. Simulation results relying on NVIDIA Isaac Sim, the underlying inferred RTAB-Map, and the advocated RL-controller are offered to provide the efficiency of the algorithm.

Keywords: Optimization, Mobile Robots, Nonlinear systems
Abstract: Dynamic feedback linearization (DFL) is a classical technique for trajectory tracking of unicycle-type mobile robots, but the resulting DFL-based controller becomes singular when the linear velocity vanishes, rendering standard DFL-based controllers unsuitable for stop-and-reverse maneuvers. This paper proposes a quadratic-program (QP)-based optimal control framework that avoids this singularity, while establishing local Lipschitz continuity of the resulting feedback law. Our approach reformulates the DFL constraints as an equality-constrained QP with a slack variable, ensuring feasibility for all states and reference signals, including at points where the robot's velocity vanishes. By introducing slack variables and tunable parameters, we demonstrate that the singular configuration can be avoided for a large class of reference trajectories. The effectiveness of the proposed approach for trajectory tracking is demonstrated through ROS2–Gazebo simulations on a TurtleBot3 Waffle robot. The code is available at https://gradslab.github.io/DFL_QP_Unicycle/

Keywords: Optimization, Reinforcement learning, Learning
Abstract: In this paper, we propose a hybrid penalty parameter selection strategy to accelerate the convergence of Consensus ADMM (C-ADMM) for distributed Multi-Robot Task Assignment (MRTA). While distributed optimization offers scalability for networked systems, the convergence speed of ADMM is known to be highly sensitive to the penalty parameter ρ. Suboptimal or static parameter choices often lead to erratic iteration counts, which is a critical limitation for time-sensitive robotic applications. To address this limitation, we introduce a finite-horizon Markov Decision Process (MDP) approach. Unlike conventional methods that rely on fixed parameters or heuristic adjustments, our proposed strategy employs a lightweight reinforcement learning policy to dynamically adapt the penalty parameter ρ during the initial transient phase based on residual trends. Subsequently, the parameter is frozen to a reliable baseline to preserve the standard fixed-parameter stability characteristics in the steady state. Numerical simulations are conducted to validate the effectiveness of the proposed method on the MURD-TAP solver. Simulation results demonstrate that the proposed approach significantly outperforms the considered baselines, achieving up to a 47.27% reduction in iterations for fully connected robot teams.

Keywords: Planning, Mobile Robots, Robotics applications
Abstract: The automation of indoor construction sites can be supported by collaborative workpiece transport with mobile robots. However, generating trajectories for such scenarios in real-time still remains challenging, since model predictive control and sampling-based methods have high computation times. This paper therefore presents a fast sampling-based planner for collaborative transport of rigid workpieces. Collision-free configurations are sampled and connected using bidirectional Informed RRT* (BI-RRT*). Bézier splines interpolate between these samples to satisfy the motion equations and limit constraint violations. For synchronous execution of paths of different lengths, we introduce a time parametrization scheme that decomposes the coupled problem into subproblems solved with S-curve profiles and quadratic programs (QP). Thus feasible trajectories are generated in under 100 ms and converge to the optimum within seconds for the tested scenarios.

Keywords:
Abstract: The scanning tunneling microscope (STM), a Nobel Prize–winning instrument, has transformed nanotechnology by enabling atomic-scale imaging, spectroscopy, and—most critically for emerging quantum technologies—atomically precise lithography on silicon for the fabrication of solid-state quantum devices.

Despite its profound impact, the fundamental feedback control loop governing STM operation has remained largely unchanged for more than four decades. This talk presents recent and ongoing efforts to fundamentally rethink STM control architecture. We examine the limitations of the conventional constant-current control paradigm and introduce new modes of operation based on alternative feedback configurations that leverage different measurement signals for closed-loop control. These approaches provide improved robustness, enhanced stability, and superior performance across a broad range of operating conditions. The practical implications of these advances are demonstrated through applications in atomic-precision lithography, high-resolution imaging, and spectroscopy. Taken together, these results suggest that revisiting STM control through the lens of modern control theory—supported by careful experimental validation—can unlock new capabilities and accelerate progress in nanoscale science, nanotechnology, and quantum device fabrication.