Racing Policy
The agent completed its first full lap after six minutes of real interaction. Peak reported performance was reached after 11 minutes of real-world experience. The final policy used controlled slipping to handle fast corners. The racing pol…
1 sources - 5 claims
The agent completed its first full lap after six minutes of real interaction. Peak reported performance was reached after 11 minutes of real-world experience. The final policy used controlled slipping to handle fast corners. The racing policy directly commanded motor torques. The task was to learn a policy that drives the Mini Wheelbot around a fixed physical track as quickly as possible.