We use reinforcement learning in simulation to obtain a driving system controlling a full-size real-world vehicle, leveraging mostly synthetic data and achieving successful sim-to-real policy transfer.
