InertiaLab

InertiaLab is a desktop + web app that turns real robot telemetry into usable rigid-body dynamics parameters (mass, COM, inertia, friction, actuator constants) without weeks of manual system ID. You import ROS bag/CSV logs, define the kinematic tree (URDF import), and select excitation segments. The app runs constrained optimization to fit parameters against measured joint torques/IMU/FT data, then generates a validated dynamics model you can export to MuJoCo/Pinocchio formats. It includes sanity checks that catch common data issues (time sync drift, encoder wrap, IMU bias, torque saturation) and reports confidence intervals so engineers know what not to trust. This is not magic: it won’t fix unobservable parameters or bad sensors, but it will drastically shorten the “why doesn’t my controller match reality?” loop for teams doing manipulation, legged robots, or cobots.

← Back to idea list