This repository contains example scripts for training reinforcement learning agents in robosuite environments using PyTorch.
humanoid.py: PPO training loop for a composite humanoid robot in robosuite using a whole-body IK controller.kinova_pick.py: TD3 training loop for a custom Kinova3 robot on the Pick environment.alert_spot.py: PPO training loop for a custom legged robot ("AlertSpot") combining Spot base and Kinova3 arm.networks.py: Implementation of TD3, PPO, and supporting neural network classes (actor, critic, replay buffer).humanoid.json: Example composite controller configuration for a humanoid robot using whole-body IK.
Install dependencies (Python 3.10+ recommended):
pip install robosuite robosuite_models torch numpy mujoco| File | Description |
|---|---|
humanoid.py |
PPO training loop for humanoid robot with whole-body IK controller. |
kinova_pick.py |
TD3 training loop for Kinova3 robot on Pick environment, with custom robot registration. |
alert_spot.py |
PPO training loop for custom legged robot combining Spot base and Kinova3 arm. |
networks.py |
Neural network implementations for TD3 and PPO agents, plus replay buffer. |
humanoid.json |
Example composite controller config for humanoid robot (whole-body IK). |
-
Copy the kinova3_6dof folder to
~/.local/lib/python3.10/site-packages/robosuite/models/assets/robots/ -
Copy the controller configs to
~/.local/lib/python3.10/site-packages/robosuite/controllers/config/robots/ -
Copy Robot class script to
~/.local/lib/python3.10/site-packages/robosuite/models/robots/manipulators/ -
Add the following import statement to
~/.local/lib/python3.10/site-packages/robosuite/models/robots/manipulators/__init__.py:
from .kinova3_6dof_robot import Kinova3_6DOF- Add the new robot to robot class mapping in
~/.local/lib/python3.10/site-packages/robosuite/robots/__init__.py
"Kinova3_6DOF": FixedBaseRobot,