Drake
MIT robotics toolbox for model-based dynamics, optimization, and control system design
License
BSD-3-Clause
Overview
Drake is a C++ and Python toolbox from the Robot Locomotion Group at MIT for model-based design and verification of robots. It provides multibody dynamics with contact modeling, mathematical optimization and a systems framework for control design. It is particularly suited to contact-rich manipulation tasks such as grasping and assembly.
Drake is a C++/Python toolbox for analyzing the dynamics of robots and building control systems, developed by the Robot Locomotion Group at MIT. It's particularly strong for contact-rich manipulation tasks.
Key features include: - Multibody dynamics with contact modeling - Mathematical programming (optimization) - Systems framework for control design - Perception pipeline integration - Visualization tools (Meshcat, Meldis) - Python and C++ interfaces - Integration with LCM for communication - Extensive manipulation examples
Drake is used by leading robotics labs for research in manipulation, locomotion, and control. It excels at tasks requiring accurate contact dynamics like grasping and assembly.
Why it matters
Drake is MIT's answer to rigorous robot design — it provides the mathematical tools to not just simulate robots, but to formally verify that control algorithms work correctly. Used by Toyota Research Institute and Boston Dynamics, it bridges the gap between academic control theory and real robot deployment.
Problem it solves
Building reliable robot controllers requires more than simulation — it requires mathematical tools to design, optimize, and verify control algorithms. Drake provides these tools in a unified framework used by leading robotics organizations.
Key takeaways
- Model-based design toolbox for rigorous robot engineering
- Advanced multibody dynamics with automatic differentiation
- Mathematical optimization tools for trajectory planning and control
- Formal verification capabilities for control system design
- Used by Toyota Research Institute and major robotics labs
Who it is for
- Control systems researchers and engineers
- Teams designing manipulation and locomotion controllers
- Robotics companies needing rigorous algorithm verification
- Graduate students studying modern robot control theory
Limitations
- Steep learning curve requiring strong math background
- C++ API is powerful but complex for simple tasks
- Smaller community than ROS-based tools
- Documentation assumes familiarity with optimization and control theory