Demonstration Code for Unitree Go2 Quadruped Robot Walking
dreamwaq (Main) • gym_basic • unitree_rl_gym • wtw
go2_walk/
├── dreamwaq/ # [Main] Rough terrain with state estimation
│ ├── legged_gym/
│ └── rsl_rl/
├── gym_basic/ # [Sub] Basic IsaacGym examples
├── unitree_rl_gym/ # [Ref] Flat terrain locomotion
└── wtw/ # [Ref] Sim-to-Real deployment
├── go2_gym/
├── go2_gym_learn/
└── go2_gym_deploy/
Rough Terrain Locomotion with State Estimation
| Author | Jungyeon Lee (curieuxjy) |
| Paper | DreamWaQ: Learning Robust Quadrupedal Locomotion |
| Achievement | 1st Place at ICRA 2023 Autonomous Quadruped Robot Challenge (QRC) Final |
| Related | Fall Recovery Task |
| Training Log | WandB Dashboard |
I independently implemented the DreamWaQ algorithm based on the paper. The core component, Context-aided Estimator Network (CENet), has been carefully implemented and verified to work as described. Feel free to explore the code and experiment with it!
📖 For detailed setup instructions including Docker configuration and training commands, please refer to the dreamwaq/README.md.
Go2 in IsaacGym
- IsaacGym ver.4 basic examples
- Camera and joint inspection demos
Flat Terrain Locomotion (Reference)
| Original | unitreerobotics/unitree_rl_gym |
| Based on | leggedrobotics/legged_gym |
Sim-to-Real Deployment (Reference)
| Paper | Walk These Ways |
| Original | Teddy-Liao/walk-these-ways-go2 |
| Based on | Improbable-AI/walk-these-ways |