Supercharged Isaac Gym environments with multi-agent and multi-algorithm support

• Isaac Gym
• IsaacGymEnvs

Optional:

• legged_gym
• rsl_rl
• rl_games
• tianshou
• stable_baseline3

cd IsaacGymMultiAgent
pip3 install -e .

Run IsaacGymEnvs experiments (rl_games PPO implementation)

cd examples/original
python3 train.py task=Ant

Run legged_gym experiments (rsl_rl PPO implementation)

cd examples/rsl_rl
python3 train.py task=a1

Run legged_gym environments with rl_games PPO

cd examples/original
python3 train.py task=LeggedGymA1

Run IsaacGym environments with rsl_rl PPO

cd examples/rsl_rl
python3 train.py task=Ant

cd examples/original
python3 train.py task=Ant test=True headless=False checkpoint=path/to/checkpoint.pt

Supported displays:

• VNC: Tested on tigervnc-standalone-server
• Alternatives: Xvfb, VirtualGL, etc.

Add headless=False to enable rendering

export DISPLAY=:1
python train.py ... headless=False

ImportError: libpython3.7m.so.1.0: cannot open shared object file: No such file or directory

Add the python library path to LD_LIBRARY_PATH

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/python/lib