isaacgymEnv_symmetry_humanoid

Supporting code and data for the paper "Motor Synergy and Energy Efficiency Emerge in Whole-body Locomotion Learning"

Install dependencies

• simulator: Isaac Gym https://developer.nvidia.com/isaac-gym
• Ubuntu 18.04 or 20.04

Download this repository and run:

pip install -e .

Training Agent

Training the humanoid walking and running task by:

python train.py task=HumanoidSymmetryWalking headless=True ++train.params.config.symmetry_index=50
python train.py task=HumanoidSymmetryrunning headless=True ++train.params.config.symmetry_index=50

The weight of symmetry loss can be adjusted by changing the value of ++train.params.config.symmetry_index.

Training data is recorded in Tensorboard

tensorboard --logdir ./runs/

Reference repository

IsaacGymEnvs: https://github.com/NVIDIA-Omniverse/IsaacGymEnvs

SynergyTool: https://github.com/JiazhengChai/synergyTool