A Simple and Effective Learning Approach to Motion Error Correction of an Industrial Robot

This paper presents a novel approach to the correction of robot steady state motion error. This approach acquires the motion error information from robot internal encoders and feeds the acquired information to an iterative learning controller (ILC) to compute a compensation variable. It then adds the compensation variable to the original position reference command of the robot to correct the steady state motion error. The proposed approach does not require any external measurement device and any changes to robot’s internal controller, therefore it can be easily implemented and embedded into a robot’s user program. The proposed approach has a nice mathematical property that guarantees its convergence. Simulation and test facility experiments have shown that the proposed approach is effective and can significantly and consistently reduce SMPE (servo motion positional error) to within ±0.05 mm, which is an 82% reduction of SMPE.