Zero-Shot MARL Benchmark in the Cyber-Physical Mobility Lab
It provides a systematic, open-source platform for analyzing sim-to-real challenges in MARL for autonomous driving, addressing the need for reproducible evaluation.
The paper presents a reproducible benchmark for evaluating sim-to-real transfer of MARL policies for CAVs, and demonstrates that deploying a SigmaRL-trained policy reveals performance degradation from both architectural differences and the sim-to-real gap.
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.