A closed-loop kinematic approach to Self-Calibration of joint offset for multi-branched robots

Abstract

Kinematic accuracy is fundamental for the performance of multi-branched robots, which are composed of multiple branches that collaborate to manipulate. Among those various factors affecting kinematic accuracy, the joint offset is the most significant contributor due to misalignment in component assembly. In this work, we present a general, fast, and low-cost approach based on closed-loop kinematics to calibrate joint offsets of multi-branched robots. A dedicated calibration board connects the distal ends of branches to form closed-loop configurations. The identification model is derived based on the closed-loop kinematic differential model. The observable indexes of multiple closed-loop are integrated to optimize the calibration board’s poses, and joint offsets are identified simultaneously. To enhance the robustness of the calibration outcomes against joint backlash noise, the Kolmogornov-Smirnov(K-S) test is performed on multiple sets of identification results to verify that joint offset obeys normal distribution. Experiments on a quadruped showed significantly improved foot-tip position accuracy.