Multi-vehicle Conflict Resolution in Highly Constrained Spaces by Merging Optimal Control and Reinforcement Learning

We present a novel method in this work to address the problem of multi-vehicle conflict resolution in highly constrained spaces. A high-fidelity optimal control problem is formulated to incorporate nonlinear, non-holonomic vehicle dynamics and exact collision avoidance constraints. Despite being high-dimensional and non-convex, we can obtain an optimal solution by learning configuration strategies with reinforcement learning (RL) in a simplified discrete environment and approaching high-quality initial guesses progressively. The simulation results show that our method can explore efficient actions to resolve conflicts in confined space and generate dexterous maneuvers that are both collision-free and kinematically feasible.