Quadruped Fall Recovery with Deep Reinforcement Learning

This repository contains the official code and simulation environment for the paper Self-Recovery of Quadrupedal Robot Using Deep Reinforcement Learning

This work proposes a DRL-based self-recovery controller for the Unitree A1 robot using Proximal Policy Optimization (PPO). The policy learns to recover from arbitrary fallen states without any predefined kinematic motion patterns or inverse kinematics.

quad-fall-recovery-rl/
│
├── quad/
│   ├── __init__.py
│   ├── envs/
│   │   ├── __init__.py
│   │   ├── quad_env.py
│   │   ├── reward.py
│   │   └── utils.py
│   │
│   └── data/
├── config.py
├── train.py
├── test.py

python train.py

Configure hyperparameters in config.py.

python test.py

If you use this repository, please cite:

Self-Recovery of Quadrupedal Robot Using Deep Reinforcement Learning
Yusuf Eren Kılıç, Yunus Emre Akar, Hakan Temeltaş, Ecem Sümer
2024 4th International Conference on Robotics, Automation and Artificial Intelligence (RAAI)
DOI: 10.1109/RAAI64504.2024.10949561