Installation Guide

① Install ISAAC SIM and ISAAC LAB

The required version of ISAAC SIM is 4.5.0. From this point on, it is assumed that you have created a virtual environment named env_isaaclab using pyenv and will be working within that environment.

② Install openarm-isaaclab

Navigate to the same directory where IsaacLab is cloned and execute the following commands:

git clone https://github.com/reazon-research/openarm_isaaclab.git
cd openarm_isaaclab
python -m pip install -e exts/openarm_isaaclab

After completing this step, the directory structure should look like this:

|-IsaacLab
|-openarm_isaaclab

③ Edit train.py and play.py

When you clone IsaacLab, the scripts train.py and play.py will be located in:

IsaacLab/scripts/reinforcement_learning/rsl_rl/

These scripts are used for training reinforcement learning models and replaying trained models. However, to recognize the gym environment of openarm_isaaclab, you need to import openarm_isaaclab at the beginning of each script.

Specifically, add the following line at the beginning of train.py and play.py:

import openarm_isaaclab.tasks.manipulation

This ensures that the environment is properly registered and recognized.

④ Run Training

Run the following command to start training:

(env_isaaclab) reazon@reazon:~/isaac/openarm_isaaclab$ bash ../IsaacLab/isaaclab.sh -p ../IsaacLab/scripts/reinforcement_learning/rsl_rl/train.py --task Isaac-Reach-OpenArm-v0 --num_envs 2048

⑤ Replay Trained Model

To replay a trained model, use the following command:

(env_isaaclab) reazon@reazon:~/isaac/openarm_isaaclab$ bash ../IsaacLab/isaaclab.sh -p ../IsaacLab/scripts/reinforcement_learning/rsl_rl/play.py --task Isaac-Reach-OpenArm-v0 --num_envs 2048