Reinforcement Learning and Distributed Model Predictive Control for Conflict Resolution in Highly Constrained Spaces