multiagent-particle-envs
multiagent-particle-envs copied to clipboard
how to fix the Observation for obstacle landmark
Hi, i created some landmarks as obstacles in my scenario like simple_tag.py, but i found that the default observation function to entity_pos only record landmarks' center positions, without any info about its radius/size, and thus in some case, when the agent, obstacle and goal_landmark are lined, the agent hesitates and cannot decide to bypass the obstacle from left or right, and keep colliding with the obstacle, is there something i misunderstand or how could i fix it ?
if there is somebody know how to deal with it, please help me, thanks for your reading and help!