Skip to content

Latest commit

 

History

History
30 lines (25 loc) · 987 Bytes

README.md

File metadata and controls

30 lines (25 loc) · 987 Bytes
Raw

mujoco-random-env

Required Version

gym == 0.23.1
mujoco_py == 2.1.2.14
mujoco == 2.2.2

You can randomly change model's parameter (e.g. agent's mass, friction, ...) without any library.

You just need to insert randomness properties when you called the envrionment.

To randomly change the agent's mass,

  env = AntRandomEnvClass(rand_mass=[1,5], rand_fric=None, render_mode=None, VERBOSE=True)

To randomly change the friction together,

  env = AntRandomEnvClass(rand_mass=[1,5], rand_fric=[0.2, 1.0], render_mode=None, VERBOSE=True)

I also added the environment with the box to make available to apply meta reinforcement learning conveniently. 
  env = AntRandomEnvClassWithBox(rand_mass_box=[1,5], rand_fric=None, render_mode=None, VERBOSE=True)

The agent with the box looks like the following image.

image