Event
Proposes an analytical robot self-modeling method
1 reports · 1 sources
Xu Chen and Ning Tan propose an analytical robot self-modeling method: appearance and rigid link assignment are obtained from multi-view images and joint readings, then metric screw axes are identified using sparse end-effector poses, yielding extrapolatable closed-form forward kinematics and an analytical Jacobian, without needing deformation networks or URDF/CAD templates. The method was validated on a seven-axis simulated robotic arm, a twelve-joint simulated quadruped, and a six-axis real robotic arm: in simulation it achieved millimeter-level grasping and closed-loop quadruped walking; on the real robot, the median error in predicting end-effector position was 6.4 mm, compared with 11.7 mm for a calibrated nominal model, and it supports reaching tasks with feedback from external markers.
Report timeline
- Analytic Robot Self-Model: Learning Screw-Axis Kinematics for Rendering and Model-Based Control
The paper proposes an explicit analytic robot self-model that requires neither deformation networks nor URDF/CAD templates. From multi-view images and joint readings, it obtains appearance and rigid link assignments, and independently identifies metric screw axes using sparse end-effector poses, yielding closed-form forward kinematics and an analytic Jacobian that can be extrapolated.
arXiv cs.RO ·