Zero-Shot MARL Benchmark in the Cyber-Physical Mobility Lab

We present a reproducible benchmark for evaluating sim-to-real transfer of Multi-Agent Reinforcement Learning (MARL) policies for Connected and Automated Vehicles (CAVs). The platform, based on the Cyber-Physical Mobility Lab (CPM Lab) [1], integrates simulation, a high-fidelity digital twin, and a physical testbed, enabling structured zero-shot evaluation of MARL motion-planning policies. We demonstrate its use by deploying a SigmaRL-trained policy [2] across all three domains, revealing two complementary sources of performance degradation: architectural differences between simulation and hardware control stacks, and the sim-to-real gap induced by increasing environmental realism. The open-source setup enables systematic analysis of sim-to-real challenges in MARL under realistic, reproducible conditions.

Less is More: Contextual Sampling for Nonlinear Data-Enabled Predictive Control

Data-enabled Predictive Control (DeePC) is a powerful data-driven approach for predictive control without requiring an explicit system model. However, its high computational cost limits its applicability to real-time robotic systems. For robotic applications such as motion planning and trajectory tracking, real-time control is crucial. Nonlinear DeePC either relies on large datasets or learning the nonlinearities to ensure predictive accuracy, leading to high computational complexity. This work introduces contextual sampling, a novel data selection strategy to handle nonlinearities for DeePC by dynamically selecting the most relevant data at each time step. By reducing the dataset size while preserving prediction accuracy, our method improves computational efficiency, of DeePC for real-time robotic applications. We validate our approach for autonomous vehicle motion planning. For a dataset size of 100 sub-trajectories, Contextual sampling DeePC reduces tracking error by 53.2 % compared to Leverage Score sampling. Additionally, Contextual sampling reduces max computation time by 87.2 % compared to using the full dataset of 491 sub-trajectories while achieving comparable tracking performance. These results highlight the potential of Contextual sampling to enable real-time, data-driven control for robotic systems.

Synchronization-Based Cooperative Distributed Model Predictive Control

Distributed control algorithms are known to reduce overall computation time compared to centralized control algorithms. However, they can result in inconsistent solutions leading to the violation of safety-critical constraints. Inconsistent solutions can arise when two or more agents compute concurrently while making predictions on each others control actions. To address this issue, we propose an iterative algorithm called Synchronization-Based Cooperative Distributed Model Predictive Control, which we presented in [1]. The algorithm consists of two steps: 1. computing the optimal control inputs for each agent and 2. synchronizing the predicted states across all agents. We demonstrate the efficacy of our algorithm in the control of multiple small-scale vehicles in our Cyber-Physical Mobility Lab.