Transferring Autonomous Driving Knowledge on Simulated and Real Intersections
This addresses the challenge of generalization in autonomous driving for intersection handling, but it is incremental as it builds on existing transfer learning methods.
The study tackled the problem of autonomous vehicles handling different intersections by showing that pre-training on one intersection and fine-tuning on another improves performance on the new task, with benefits transferring from simulation to real vehicles, achieving better results compared to training in isolation.
We view intersection handling on autonomous vehicles as a reinforcement learning problem, and study its behavior in a transfer learning setting. We show that a network trained on one type of intersection generally is not able to generalize to other intersections. However, a network that is pre-trained on one intersection and fine-tuned on another performs better on the new task compared to training in isolation. This network also retains knowledge of the prior task, even though some forgetting occurs. Finally, we show that the benefits of fine-tuning hold when transferring simulated intersection handling knowledge to a real autonomous vehicle.