Decentralized Autonomous Traffic Management through Corridor Networks

The article presents a decentralized traffic management approach for Advanced Air Mobility (AAM) corridors using multi-agent reinforcement learning (MARL). The developed policies effectively manage traffic in complex multi-corridor environments, demonstrating strong zero-shot transfer across varying traffic densities and network geometries without centralized coordination. This advancement is significant for practitioners as it offers a scalable solution to coordinate high-density autonomous aircraft operations, enhancing flexibility in trajectory planning while ensuring safety and efficiency.