Mujin Inc. Internship - Dynamics Identification for Industrial Robotics

During my internship at Mujin Inc. in Tokyo, Japan, I developed a multi-purpose robot dynamics identification feature for the Mujin Controller, "The world's first motion planning AI-equipped intelligent robot controller".

Friction Test Results with Friction Coefficient Estimates

The dynamics identification feature improves the performance of the robot motion planning by improving the robot torque model. The feature estimates the coefficients for:

- Motor Friction (viscous, coulomb)

- Mass & COM of Links

- Inertia Tensor

---

The results are visualized in simulation using experimental results. The video above shows automated inertia pose validation.

The feature was fully developed into production code and is now shipped on Mujin controllers for use by industry leaders such as PALTAC, Askul, JD.com, and others

COM visualization dev tool, shows Mass*Center of Mass optimization possibilities

During this project, I also completed:

- Integrated Mujin's automated robot test trajectory generation

- Developed proprietary data analysis, fitting, and optimization tools & algorithms

- Refined UX/UI data visualization tools unique for dynamics validation

Short cameo of me working on dynamics identification feature @ 0:33 :)