Georg Schildbach

Max Studt, Georg Schildbach

Decentralized collision avoidance remains challenging, particularly when agents do not communicate any information related to planned trajectories. Most existing approaches either rely on conservative coordination mechanisms or provide limited guarantees on recursive feasibility and convergence. This paper develops a decentralized contingency MPC framework for multi-agent systems with nonlinear dynamics that achieves collision-free motion under a state-only information pattern. Each agent follows the same consensual rule set, enabling safe decentralized planning without communication. Each agent solves a local optimization problem that couples a nominal trajectory with a contingency certificate ensuring a feasible backup maneuver under receding-horizon operation. A novel geometric and decentralized safe-set update mechanism prevents feasibility loss between consecutive time steps. The resulting scheme guarantees recursive feasibility, including collision avoidance, and establishes a Lyapunov-type convergence result to an admissible safe equilibrium. Simulation results demonstrate performance in both sparse and dense multi-agent environments, including cluttered bottleneck scenarios and under plug-and-play operation.

Fabian Domberg, Georg Schildbach

As learning-based robotic controllers are typically trained offline and deployed with fixed parameters, their ability to cope with unforeseen changes during operation is limited. Biologically inspired, this work presents a framework for online Continual Reinforcement Learning that enables automated adaptation during deployment. Building on DreamerV3, a model-based Reinforcement Learning algorithm, the proposed method leverages world model prediction residuals to detect out-of-distribution events and automatically trigger finetuning. Adaptation progress is monitored using both task-level performance signals and internal training metrics, allowing convergence to be assessed without external supervision and domain knowledge. The approach is validated on a variety of contemporary continuous control problems, including a quadruped robot in high-fidelity simulation, and a real-world model vehicle. Relevant metrics and their interpretation are presented and discussed, as well as resulting trade-offs described. The results sketch out how autonomous robotic agents could once move beyond static training regimes toward adaptive systems capable of self-reflection and -improvement during operation, just like their biological counterparts.

Michael Kurenkov, Sajad Marvi, Julian Schmidt et al.

The increasing interest in autonomous driving systems has highlighted the need for an in-depth analysis of human driving behavior in diverse scenarios. Analyzing human data is crucial for developing autonomous systems that replicate safe driving practices and ensure seamless integration into human-dominated environments. This paper presents a comparative evaluation of human compliance with traffic and safety rules across multiple trajectory prediction datasets, including Argoverse 2, nuPlan, Lyft, and DeepUrban. By defining and leveraging existing safety and behavior-related metrics, such as time to collision, adherence to speed limits, and interactions with other traffic participants, we aim to provide a comprehensive understanding of each datasets strengths and limitations. Our analysis focuses on the distribution of data samples, identifying noise, outliers, and undesirable behaviors exhibited by human drivers in both the training and validation sets. The results underscore the need for applying robust filtering techniques to certain datasets due to high levels of noise and the presence of such undesirable behaviors.

Fabian Domberg, Georg Schildbach

Learning-based controllers are often purposefully kept out of real-world applications due to concerns about their safety and reliability. We explore how state-of-the-art world models in Model-Based Reinforcement Learning can be utilized beyond the training phase to ensure a deployed policy only operates within regions of the state-space it is sufficiently familiar with. This is achieved by continuously monitoring discrepancies between a world model's predictions and observed system behavior during inference. It allows for triggering appropriate measures, such as an emergency stop, once an error threshold is surpassed. This does not require any task-specific knowledge and is thus universally applicable. Simulated experiments on established robot control tasks show the effectiveness of this method, recognizing changes in local robot geometry and global gravitational magnitude. Real-world experiments using an agile quadcopter further demonstrate the benefits of this approach by detecting unexpected forces acting on the vehicle. These results indicate how even in new and adverse conditions, safe and reliable operation of otherwise unpredictable learning-based controllers can be achieved.

Honghu Xue, Benedikt Hein, Mohamed Bakr et al.

We propose a deep reinforcement learning approach for solving a mapless navigation problem in warehouse scenarios. In our approach, an automation guided vehicle is equipped with LiDAR and frontal RGB sensors and learns to perform a targeted navigation task. The challenges reside in the sparseness of positive samples for learning, multi-modal sensor perception with partial observability, the demand for accurate steering maneuvers together with long training cycles. To address these points, we propose NavACL-Q as a method for automatic curriculum learning in combination with a distributed version of the soft actor-critic algorithm. The performance of the learning algorithm is evaluated exhaustively in an unseen warehouse environment to validate both robustness and generalizability of the learned policy. Results in NVIDIA Isaac Sim demonstrates that our trained agent significantly outperforms a map-based navigation pipeline provided by NVIDIA Isaac Sim with an increased agent-goal distance of 3m and wider initial relative agent-goal rotations of 45 degree. The ablation studies also suggests that NavACL-Q greatly facilitates the learning process with a performance gain of roughly 40% compared to training with random starts and that the utilization of a pre-trained feature extractor manifestly boosts the performance by approximately 60%.