mdp_playground.envs.mujoco_env_wrapper.get_mujoco_wrapper¶
-
mdp_playground.envs.mujoco_env_wrapper.get_mujoco_wrapper(base_class)[source]¶ Wraps a mujoco-py environment to be able to modify its XML attributes and inject the dimensions from MDP Playground. The values for these dimensions are passed in a config dict as for mdp_playground.envs.RLToyEnv.
- Currently supported dimensions:
time_unit action_space_max
For both of these dimensions the scalar value passed in the dict is used to multiply the base environments values. For the Mujoco envs, the time_unit is achieved by multiplying the Gym Mujoco env’s frame_skip and thus needs to be such that time_unit * frame_skip is an integer. The time_unit is NOT achieved by changing Mujoco’s timestep because that would change the numerical integration done my Mujoco.