This paper presents a cooperative distributed model predictive control (MPC) scheme for nonlinear continuous-time systems. The centralized optimal control problem is solved asynchronously via a fixed number of sensitivity-based distributed programming (SBDP) iterations. The proposed scheme requires only neighbor-to-neighbor communication and no synchronization between agents during optimization. Under nominal MPC stability and bounded information delay, local exponential stability is established for a sufficiently large number of per-agent SBDP iterations. Numerical and hardware-in-the-loop results on both Ethernet and Wi-Fi demonstrate the benefits of an asynchronous execution, reducing execution times by over 60

Automated driving (AD) is promising, but the transition to fully autonomous driving is, among other things, subject to the real, ever-changing open world and the resulting challenges. However, research in the field of AD demonstrates the ability of artificial intelligence (AI) to outperform classical approaches, handle higher complexities, and reach a new level of autonomy. At the same time, the use of AI raises further questions of safety and transferability. To identify the challenges and opportunities arising from AI concerning autonomous driving functionalities, we have analyzed the current state of AD, outlined limitations, and identified foreseeable technological possibilities. Thereby, various further challenges are examined in the context of prospective developments. In this way, this article reconsiders fully autonomous driving with respect to advancements in the field of AI and carves out the respective needs and resulting research questions.

Vision-language-action (VLA) driving models couple a reasoning stage with a diffusion-based trajectory decoder, but do not give a direct way to redirect attention toward safety-critical actors at inference time without retraining. We studied a bounded additive pre-softmax attention bias on the visual tokens of detector localized traffic actors on Alpamayo-R1's Qwen3-VL backbone. It is applied as a fail open forward pre-hook with no weight changes. On 50 lane-change scenarios from the Physical AI World Model Synthetic dataset. The trajectory decoder shows a monotonic dose response in the bias magnitude, separate from a paired zero bias control at every tested magnitude. It reaches ≈ 17 cm mean displacement with lateral shifts up to ∼ 140 cm at the clamp. A layer ablation places the action-relevant signal in late layers, where the effect increases with the number of hooked layers (2.0cm for the first 8 layers; 67.6cm for all 36). A per call injection audit explains why the Chain-of-Causation text never changes. The mask based bias never reaches the reasoning pathway in this serving stack, so the invariance is verified exposure, not robustness. Steered trajectories tend to shift toward the attended actor, suggesting the bias governs where the model looks rather than encoding a target behavior.

With the growing importance of voltage source converters in the electrical power grid, ensuring their reliability by fulfilling performance requirements is becoming increasingly important. At the same time, the passivity of the converter input admittance is crucial to ensure the stability of the interaction with the AC grid. Fulfilling all requirements at the same time renders the design of the AC current control (ACCC) a challenging task. This paper proposes a pole placement method that incorporates an explicit passivity constraint for the ACCC of modular multilevel converters in HVDC applications. Based on a simplified analytical model, the design parameters are derived directly from the performance specifications. Subsequently, the remaining degrees of freedom are utilized to fulfill the passivity requirement, making the underlying constraints and trade-offs transparent. The method is validated through MATLAB simulations using a representative design example. A comparison with an H∞-based optimization method from the literature demonstrates that the proposed approach yields competitive performance with a lower control order. Finally, a robustness analysis is conducted to assess the effect of previously neglected model dynamics and the grid model. The results show that delays in the automation interface have a small impact on all objectives except passivity, whose sensitivity depends mainly on the delay magnitude and the desired tracking speed.

This paper presents a novel sensitivity-based distributed programming (SBDP) approach for non-convex, large-scale nonlinear programs (NLP). The algorithm relies on first-order sensitivities to cooperatively solve the central NLP in a distributed manner with only neighbor-to-neighbor communication and parallelizable local computations. The decoupling of the subsystems is based on primal decomposition. We derive sufficient local convergence conditions for non-convex problems. Furthermore, we consider the SBDP method in a distributed optimal control context and derive favorable convergence properties in this setting. We illustrate these theoretical findings and the performance of the proposed method with a comparison to state-of-the-art algorithms and simulations of various distributed optimization and control problems.

This paper presents a concise overview of sensitivity-based methods for solving large-scale optimization problems in distributed fashion. The approach relies on sensitivities and primal decomposition to achieve coordination between the subsystems while requiring only local computations with neighbor-to-neighbor communication. We give a brief historical synopsis of its development and apply it to both static and dynamic optimization problems. Furthermore, a real-time capable distributed model predictive controller is proposed which is experimentally validated on a coupled water tank system. Dieser Beitrag bietet einen kompakten & Uuml;berblick zu sensitivit & auml;tsbasierten Verfahren f & uuml;r die verteilte L & ouml;sung hochdimensionaler Optimierungsprobleme. Das Schema nutzt Sensitivit & auml;ten und primale Dekomposition, um die Koordination zwischen den Teilsystemen sicherzustellen, wobei lediglich lokale Berechnungen sowie die Kommunikation mit den jeweiligen Nachbarn erforderlich sind. Zus & auml;tzlich zu einer historischen Einordnung wird sowohl die Anwendung auf statische als auch auf dynamische Optimierungsprobleme betrachtet. Dar & uuml;ber hinaus wird ein echtzeitf & auml;higes verteiltes modellpr & auml;diktives Regelungsverfahren vorgestellt, welches experimentell an einem gekoppelten Wassertanksystem validiert wird.

Data-based learning of system dynamics allows model-based control approaches to be applied to systems with partially unknown dynamics. Gaussian process regression is a preferred approach that outputs not only the learned system model but also the variance of the model, which can be seen as a measure of uncertainty. Stochastic model predictive control uses a stochastic system model like this to propagate the probability density functions of the predicted states and solves a stochastic optimal control problem in every time step. This paper proposes a stochastic model predictive control algorithm with guarantees on recursive feasibility and stability for nonlinear systems with unknown parts that are learned from data using Gaussian process regression. To this end, an error bound of the Gaussian process prediction is established based on an upper bound of the norm of the unknown dynamics function in the corresponding reproducing kernel Hilbert space. Unlike related work, this allows to consider state and input dependent disturbances instead of independent and identically distributed (iid) white Gaussian noise which is usually considered in the context of stochastic model predictive control. The numerical evaluation illustrates that the resulting control algorithm stabilizes nonlinear systems with partially unknown dynamics.

The road friction coefficient is a key parameter for traffic safety and for advanced driver-assistance systems (ADAS) such as electronic stability control and trajectory planning in autonomous vehicles. Most existing estimation methods rely on dynamic maneuvers, whereas friction estimation during steering at standstill and very low speeds has received limited attention. This paper addresses this gap by introducing two novel friction estimation approaches based on steering torque in standstill and low-speed conditions. The first method employs a Kalman-Bucy filter (KBF) coupled with a simplified brush model, using a single tread element formulation to determine when estimation should be updated or halted. The second method applies an unscented Kalman filter (UKF) to an extended torsional spring model that incorporates a damping term for low-speed maneuvers, while parametric output sensitivity (POS) analysis is used to assess identifiability. Both approaches are compared to an existing method from the literature and are validated in simulation and with experimental data from three vehicles of varying complexity. The results demonstrate accurate and consistent friction estimation during steering at standstill and slow rolling maneuvers, extending the applicability of effect-based methods beyond conventional dynamic driving scenarios.

The reliability of machinery plays an essential role in industrial practice, which includes the growing topic of fault detection for electric motors. Since permanent magnet synchronous motor (PMSMs) usually have built-in current and speed sensors, it is advantageous to use them for fault detection purposes as they enable a non-invasive and cost-effective implementation. Focusing on speed and current signals, two methods are developed for the detection of localized bearing faults in this paper. They leverage envelope analysis and the wavelet packet transform for feature extraction before classification is performed using a support vector machine. In addition, properties of vibration signals and built-in sensor signals are discussed and important similarities are highlighted. The effectiveness of the two methods is demonstrated in the analysis of experimental measurements, where non-artificial localized bearing faults were investigated by means of the phase currents, the d-current and q-current as well as the speed signal along with a comparison to vibration signal analysis. Both methods are shown to be effective for bearing fault diagnosis and exhibit higher detection accuracies than comparable approaches, with the best results being achieved using the q-current. This highlights the viability of built-in sensors in this context.

Building energy systems pose challenging control tasks when aiming for energy efficient though thermally comfortable control of heating, cooling, and ventilation. Rule based controllers can be easily outperformed by taking system predictions and weather forecasts into account. For model predictive control (MPC), the choice of the cost function is crucial. Here, we show a multi-objective MPC case study and present automated decision making approaches. This reveals the trade-off between cost functions and underlines their importance in MPC control design for building energy systems.

This paper presents a distributed model predictive control (DMPC) scheme for nonlinear continuous-time systems. The underlying distributed optimal control problem is cooperatively solved in parallel via a sensitivity-based algorithm. The algorithm is fully distributed in the sense that only one neighbor-to-neighbor communication step per iteration is necessary and that all computations are performed locally. Sufficient conditions are derived for the algorithm to converge towards the central solution. Based on this result, stability is shown for the suboptimal DMPC scheme under inexact minimization with the sensitivity-based algorithm and verified with numerical simulations. In particular, stability can be guaranteed with either a suitable stopping criterion or a fixed number of algorithm iterations in each MPC sampling step which allows for a real-time capable implementation.

Model Predictive Control (MPC) is a promising method for flight control, offering precise stabilization and maneuvering by predicting system behavior using a model of the aircraft dynamics. Essential for these dynamics are the aerodynamic coefficients While conventional aerodynamic models often do not meet the real-time requirements of flight control applications, neural networks (NN) promise to accurately capture aerodynamic behavior. However, their computational feasibility in real-time MPC remains an active research area. This paper presents a nonlinear Model Predictive Flight Control strategy for a fighter aircraft, where the numerical solution of the MPC problem requires the gradients of the aerodynamic tables. Instead of modeling the aerodynamic coefficients directly with NNs, we propose to use the original look-up tables and only model their derivatives with low-dimensional feedforward NNs. Simulation results of an MPC demonstrate enhanced computational efficiency without sacrificing accuracy, where the NN modeling makes the gradient computation more than three times faster than conventional difference quotient calculations.

This article presents a sensitivity-based algorithm for distributed optimal control problems (OCP) of multi-agent systems with nonlinear dynamics and state/input couplings, as they arise, for instance, in distributed model predictive control. The algorithm relies on first-order sensitivities to cooperatively solve the distributed OCP in parallel. The solutions to the resulting local OCPs are computed with a fixed-point scheme and communicated within one communication step per algorithm iteration to the neighbors. Convergence results are presented under the inexact minimization of the local OCP. The algorithm is evaluated in numerical simulations for an example system.

This brief presents the experimental results of a sensitivity-based distributed nonlinear model predictive (DMPC) scheme applied to a multiagent levitating planar motion system. The algorithm is based on first-order sensitivities such that the central optimal control problem (OCP) is solved cooperatively and in parallel on distributed hardware with networked communication. The experiments consist of a leader-follower scenario, a distribution problem, formation control, and cooperative load transport. The scenarios include couplings in cost functions, constraints, and dynamics in addition to inhomogeneous and nonlinear agent dynamics providing a challenging validation environment. The results showcase the applicability of DMPC to a wide range of classical distributed control problems and demonstrate the real-time capability of the proposed approach.

Most modern robotic joints are equipped with strain wave gears (SWG) as well as measurement devices like encoders and torque transducers. Torque information can also be gained without torque transducers from strain gauge sensors that are mounted on the flex spline, the deformable part of SWGs. These sensor signals provide, in addition, information about the input rotation speed and can therefore be exploited for its estimation. This paper focuses on this input rotation speed estimation by means of a peak-to-peak scheme, a trigonometric function approach by utilizing an alternative sensor setup, an extended Kalman filter (EKF), and a frequency-locked loop. The analysis leads to the conclusion that the EKF shows the highest potential to accomplish this task.

While artificial intelligence (AI) is advancing rapidly and mastering increasingly complex problems with astonishing performance, the safety assurance of such systems is a major concern. Particularly in the context of safety-critical, real-world cyber-physical systems, AI promises to achieve a new level of autonomy but is hampered by a lack of safety assurance. While data-driven control takes up recent developments in AI to improve control systems, control theory in general could be leveraged to improve AI safety. Therefore, this article outlines a new perspective on AI safety based on an interdisciplinary interpretation of the underlying data-generation process and the respective abstraction by AI systems in a system theory-inspired and system analysis-driven manner. In this context, the new perspective, also referred to as data control, aims to stimulate AI engineering to take advantage of existing safety analysis and assurance in an interdisciplinary way to drive the paradigm of data control. Following a top-down approach, a generic foundation for safety analysis and assurance is outlined at an abstract level that can be refined for specific AI systems and applications and is prepared for future innovation.

In the trajectory planning of automated driving, data-driven statistical artificial intelligence (AI) methods are increasingly established for predicting the emergent behavior of other road users. While these methods achieve exceptional performance in defined datasets, they usually rely on the independent and identically distributed (i.i.d.) assumption and thus tend to be vulnerable to distribution shifts that occur in the real world. In addition, these methods lack explainability due to their black box nature, which poses further challenges in terms of the approval process and social trustworthiness. Therefore, in order to use the capabilities of data-driven statistical AI methods in a reliable and trustworthy manner, the concept of TrustMHE is introduced and investigated in this paper. TrustMHE represents a complementary approach, independent of the underlying AI systems, that combines AI-driven out-of-distribution detection with control-driven moving horizon estimation (MHE) to enable not only detection and monitoring, but also intervention. The effectiveness of the proposed TrustMHE is evaluated and proven in three simulation scenarios.

Human-robot collision avoidance plays an essential role in facilitating the integration of robotic systems. However, many state-of-the-art approaches do not consider the future movements of dynamic obstacles or neglect the importance of allowing trajectory tracking behavior. To this end, a method based on model predictive control (MPC) and dynamic obstacle prediction is proposed that features trajectory tracking with minimal error in obstacle-free cases. Otherwise, collisions with dynamic obstacles are avoided by either only adapting the speed or by additional local deviations from the path while quickly recovering to a low tracking error afterwards. This MPC formulation is applied to an omnidirectional mobile robot within simulations and real-world experiments that demonstrate the effectiveness of the proposed approach with respect to tracking accuracy and human safety.

The geometry of a proportional electromagnetic actuator is optimized using a recent method for semi-infinite programming. To ensure cost-efficient manufacturing, both the nominal geometry and the manufacturing tolerances are optimized simultaneously to achieve the required force characteristics across all tolerance combinations. To minimize the number of FEM simulations, the applied method employs adaptive learning via a Gaussian process model to model the relationship between the nominal geometry, tolerances, and the resulting force characteristics. Moreover, the proposed algorithm avoids multiple optimization stages but allows for a direct solution of semi-infinite programs, which further increases the sample efficiency. Two formulations of the optimization problem are compared based on the number of FEM simulations required. It is demonstrated that a feasible nominal geometry and tolerances can be identified with reasonable overhead compared to optimizing the nominal geometry alone. Further, it is shown that in contrast to the simultaneous optimization the combination of nominal optimal geometries with separately defined tolerances leads to actuators which violate the specification. Copyright (c) 2025 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

This paper presents a model predictive controller for truck-trailer systems in off-road environments that takes into account uncertainties of the employed vehicle model. Model predictive controllers are widely used in the field of truck-trailer systems, as they enable to plan complex maneuvers, account for obstacles and consider system limits. A common model is derived from the kinematics of the vehicle. This model is based on the assumptions of flat surfaces and no slip. Driving in harsh environments, however, these assumptions are often violated, resulting in model uncertainties and thus in an impaired tracking accuracy. To address this issue, this paper enhances the kinematic model by Gaussian Process-based correction models that are adapted to the road conditions online. To avoid potentially dangerous maneuvers, the paper further proposes to consider the prevention of high model uncertainties an objective of the controller. The presented methods are evaluated in a high-fidelity simulation environment.