Hint

Example code: https://github.com/cselab/korali/tree/master/examples/study.cases/openAIGym/

Reinforcement Learning examples on OpenAI Gym

This folders contain a ready-to-use setup to run OpenAI Gym, both for Multi-Joint dynamics with Contact (MuJoCo) and PyBullet environments.

Pre-Requisites:

To use MuJoCo environments, follow the installation instructions here: https://github.com/openai/mujoco-py#install-mujoco. For this, you will first need to acquire a MuJoCo license To use PyBullet environments, simply run the ./install_deps.sh script. PyBullet requires no license to run.

Running an environment:

Any of the following environments are available for testing:

% MuJoCo
Ant-v2
HalfCheetah-v2
Hopper-v2
Humanoid-v2
HumanoidStandup-v2
InvertedDoublePendulum-v2
InvertedPendulum-v2
Reacher-v2
Swimmer-v2
Walker2d-v2

% PyBullet
AntBulletEnv-v0
HalfCheetahBulletEnv-v0
HopperBulletEnv-v0
HumanoidBulletEnv-v0
Walker2DBulletEnv-v0

To run any of these, use the following example:

python3 run-vracer.py --env AntBulletEnv-v0

Producing a movie:

To generate a movie that displays the outcome of a particular trained policy, use the following command:

python3 ./genMovie --env AntBulletEnv-v0 --input _result_vracer_AntBulletEnv-v0 --output myMovie

The command will read the result of training an AntBulletEnv-v0 environment from the _result_vracer_AntBulletEnv folder and output an .mp4 movie in the myMovie folder.