unusual vehicle behavior #5
Description
During evaluation, I encounter unusual vehicle behavior:
With PPO, 'RL' vehicles make hard stops even when no 'human' vehicles are following them. Their speed sometimes resets to zero at the edge before 'TL' junction, even when the edge is empty.
With DDPG, the behavior is more erratic. The first 'RL' vehicles passing the junction before the TL stop completely and remain stopped for the rest of the simulation, blocking all other vehicles since the lanes stay inaccessible. I've attached a video demonstrating these issues for clarity. Have you any thoughts?
ppo eval.zip