Dear developers,
I 'm using the commonroad rl package now. Since the reward result is not so good, I want to know how the agent really act in the scenarios, in which the visualization is needed. I want to use the commonroad io to treat the agent as a dynamic obstacle, but I need to know the state(position, velocity,orientation)in each step. How can I do that, I cannot get them directly from the observation that returned by step() function right?
BTW, maybe my way of doing it is wrong, is there any better way to visualize the policy that act by agent?
Hi, you can directly evaluate and visualize a trained agent using this script: commonroad_rl/evaluate_model_sumo.py · master · tum-cps / commonroad-rl · GitLab.
Wow, that’s amazing ! Thanks for that!