CommonRoad is an open-source software focusing on motion planning problems, where traffic scenarios can be converted from real traffic datasets (e.g., NGSIM, highD, inD…). With the development of Cooperative Adaptive Cruise Control (CACC) technology, many researchers have developed vehicle platooning toolboxes based on open-source software, while multi-vehicle interaction problems still hold strong research value. CommonRoad-MARL provides engine dynamics models and vehicle platooning functionality. By adding engine dynamics models, more realistic dynamic behaviors can be obtained, which can also be used for heterogeneous vehicle platooning research.

In real traffic and vehicle platooning, vehicles (agents) need to interact with other participants in the platoon or other vehicles. Multi-agent reinforcement learning (MARL) is closely related to such interactive environments. Through MARL, all vehicles in the platoon can achieve greater intelligence and safety potential, which aligns with future development trends. Therefore, we developed CommonRoad-MARL to support multi-agent reinforcement learning research for vehicle platooning.