An OpenAI Gym style reinforcement learning interface for Agility Robotics' biped robot Cassie
- Place
mjpro150/
andmjkey.txt
in thecassie/cassiemujoco
folder (see http://www.mujoco.org/ for info on downloading MuJoCo and obtaining a license). - Place
libcassiemujoco.so
in thecassie/cassiemujoco
folder (see https://github.com/osudrl/cassie-mujoco-sim for info on compiling libcassiemujoco.so). Temporary note: cassie-rl currently uses a different version of cassie-mujoco-sim with a foot position API. See my fork of the former. - Run
$ pip install -e .
in root folder (the one that contains setup.py).
To test installation:
$ python3
> from cassie import CassieEnv
> env = CassieEnv("walking")
> env.render()
Add the OpenAI gym environment registration code so that the environment can be created with gym.make().