Overview

This repo contains the code for learning an RL policy in IsaacGym for a two wheeled inverted pendulum (TWIP) and a Flywheel inverted pendulum robots. It includes the following folders and subfolders:

1. assets: contains the URDF file of the robot

2. code: contains the IsaacGym environment and the the task specification/PPO parameters

How to start the learning

1. Download IsaacGym

2. Copy the URDF file contained in the assets folder in IsaacGymEnvs/assets

3. Copy Twip.yaml and FlywheelPendulum.yaml in the folder in IsaacGymEnvs/isaacgymenvs/cfg/task

4. Copy TwipPPO.yaml and FlywheelPendulumPPO.yaml in the folder in IsaacGymEnvs/isaacgymenvs/cfg/train

5. Copy twip.py and flywheel_pendulum.pyin the assets folder in IsaacGymEnvs/isaacgymenvs/task

6. Paste in IsaacGymEnvs/isaacgymenvs/tasks/init.py

from .twip import Twip
from .flywheel_pendulum import FlywheelPendulum

....

"Twip": Twip,
"FlywheelPendulum": FlywheelPendulum,

7. run python train.py task=Twip or python train.py task=FlywheelPendulum