Imitation Is Not Enough: Robustifying Imitation with Reinforcement Learning for Challenging Driving Scenarios