Open-source software stack for legged robot locomotion, manipulation, and autonomy
We are a research group at Beijing Institute of Technology focused on the cerebellum and brain of legged robots — motion planning and control for bipeds, quadrupeds, humanoid manipulators, and quadruped-arm systems. Our work spans the full stack from low-level dynamics and optimization to high-level task reasoning and autonomous deployment.
What we build: Tools and frameworks covering the complete legged robotics pipeline — traditional model-based control (MPC, QP, trajectory optimization) and learning-based control (deep RL, VLA, whole-body motion), from simulation training to real-world deployment (sim2real).
Our tech stack: Deep reinforcement learning · Multi-body dynamics modeling · Optimal & modern control · Trajectory optimization · MPC · Quadratic programming · Nonlinear solvers · Vision-Language-Action (VLA) · World Action Model (WAM)
| Project | Description | Language | Stars |
|---|---|---|---|
| AStarFootstepPlanner | A* footstep planner for humanoid robots — kinematic constraints, obstacle avoidance, stepping stones | C++ |  |
| Heuclid | Lightweight Euclidean geometry library — convex hull, polygon intersection (SAT), Bézier curves | C++ |  |
| rsl-rl-ex (coming soon) | Extended RL framework for legged locomotion — based on RSL-RL with additional algorithms and environments | Python |  |
| Project | Description | Language | Stars |
|---|---|---|---|
| robot-motion-player | CLI/GUI for robot motion datasets — MuJoCo playback, IK tuning, AMP quality metrics, GIF/video export | Python |  |
| leggedlabultra (coming soon) | Training framework for legged locomotion policies — simulation environments, reward engineering, domain randomization | Python |  |
| Project | Description | Language | Stars |
|---|---|---|---|
| BitbotDeploy (coming soon) | Sim-to-real (sim2real) deployment framework for legged robots — policy export, hardware abstraction, real-time control loop | C++ / Python |  |
| Project | Description | Language | Stars |
|---|---|---|---|
| Coppeliasim-Tutorials | Bilingual (EN/CN) robotics simulation teaching kit — 6-hour curriculum, slides, models | MATLAB |  |
| OpenClaw-Guide | Deployment guide for OpenClaw AI agent gateway — platform setup, API config, 10+ messaging platforms | Shell |  |
- Gait Generation — motion trajectory generation for legged robots (joint trajectories, swing/stance patterns)
- VLA Framework — Vision-Language-Action model for legged robot task execution
- WAM Integration — World Action Model for humanoid whole-body control
Affiliated with Beijing Institute of Technology (BIT) · Built with ❤️ for the legged robotics research community
