Bridging the Gap Between Simulation and Reality

摘要

Transferring robotic control policies --- learned in simulation --- to physical robots is a promising alternative to learning directly on the physical system. Unfortunately, policies learned in simulation often fail in the real world due to the inevitable discrepancies between the real world and simulation. This thesis aims to bridge the gap between simulation and reality by developing methods for grounding simulation to reality and developing methods for assessing how well a policy learned in simulation will perform before it is executed in the real world. We discuss completed work towards a simulation-transfer method and methods of safe policy evaluation. We then present directions for future work in these areas.