A legged_gym based framework for training legged robots in Genesis and IsaacGym
-
Totally based on legged_gym
-
Integration of 2 simulators: Genesis and IsaacGym
You can train policies in two simulators using one framework and two conda environments.
-
Incorporation of various methods in published RL papers
Method Paper Link Code Periodic Gait Reward Sim-to-Real Learning of All Common Bipedal Gaits via Periodic Reward Composition go2_wtw Walk These Ways Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behavior go2_wtw System Identification Learning Agile Bipedal Motions on a Quadrupedal Robot go2_sysid Teacher-Student Rapid Locomotion via Reinforcement Learning go2_ts EstimatorNet Concurrent Training of a Control Policy and a State Estimator for Dynamic and Robust Legged Locomotion go2_ee Constraints as Terminations CaT: Constraints as Terminations for Legged Locomotion Reinforcement Learning go2_cat DreamWaQ DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning go2_dreamwaq
Please refer to the doc of this repo.
Note
- Currently(Genesis 0.3.6), compiling with DepthCamera sensors and heightfield terrain added will invoke CUDA out of memory error (50 envs, on RTX3080 10GB). Users should use GPU with larger graphics memory.
- Try to replace mesh in robot descrption files(urdf, xml and usd) with basic geometries when using Genesis. It will slow down the simulation and potentially cause NaN values in robot states(base_ang_vel, dof_pos, dof_vel)
| Robot | Sim | Real |
|---|---|---|
| Unitree Go2 | video | video |
| Limx TRON1_PF (flat) | ![]() |
video |
| TRON1_PF (rough) | ![]() |
video |
| Limx TRON1_SF (flat) |
- Add go2 deploy demos and instructions (2025/07/25)
- Add explicit estimator support (2025/09/29)
- Add teacher-student implementation (2025/08/02)
- Add Constraints as Terminations implementation (2025/10/23)
- Add support for TRON1_PF (flat and rough)
- Add DreamWaQ implementation (2025/11/27)
- Add support for TRON1_SF (flat and rough)
- Add support for TRON1_WF (flat and rough)
- Add support for warp-based depth camera

