Identifying joint clearance via robot manipulation

Abstract

Inaccuracy in robot manipulation is a result of various uncertainties. Most methods reduce operation errors by calibrating robot parameters, with little attention on understanding the uncertainty sources in the process. This paper investigates how operation accuracy of robot manipulators can be improved by identifying one of the major uncertainty–joint clearance. We first develop the dynamic model of a Delta robot with joint clearance to obtain the operation error of a given trajectory. Errors with different operating procedures can, therefore, be calculated. We then use a Kriging-based model to relate manipulator performances with joint clearance values. Real-time calibration can then be performed by identifying joint clearance via experiments. Errors can also be reduced using optimal path planning with the calibrated joint clearance. Results show that this method reduces the average error at target points from 0.637 to 0.031 mm for robot manipulators with joint clearances of 0.328, 0.171, and 0.483 mm. This is a 95.1% improvement in accuracy over that for the manipulator before optimization. The proposed method can help manufacturers determine robot quality, and achieve optimal operation in a workspace with improved accuracy.