Project
Drake
MIT robotics toolbox for modeling, planning, optimization-based control, and simulation.
ProjectAdvanced
Pinocchio
Efficient rigid-body dynamics library for modeling, control, and optimal control.
Overview
Pinocchio is a high-performance rigid-body dynamics library used for robot modeling, control, and optimal control. It is closer to classical robotics than deep learning stacks: forward/inverse dynamics, kinematics, and derivatives that feed planners and controllers.
Use it when learning-based policies must interface with model-based control, or when you need accurate dynamics inside a custom controller.
Who it is for
Control engineers; model-based robotics; hybrid learning+control stacks.
Key highlights
- Fast rigid-body dynamics
- Python and C++ APIs
- Standard in model-based robotics
When to use
- IDC/IK, dynamics, trajectory optimization backends
When not to use
- End-to-end VLA training as the only goal
Getting started
- 1Install pinocchio
- 2Load a URDF and compute FK/ID
- 3Integrate with your controller loop
Caveats & pitfalls
- Steeper learning curve if you only know PyTorch.
Content reviewed 2026-07-23