Inverse Reinforcement Learning for Collision Avoidance and Trajectory Prediction in Distributed Reconfigurations