Adaptive Robotic Deburring

Abstract

The paper describes the development of an active end effector force control system for robotic deburring. The purpose of the control is to maintain a constant cutting force by making position adjustments normal to the part edge. Autoregressive moving-average (ARMA) models which relate the position command for the end effector to the resulting change in the cutting force were developed for the deburring process. A non-adaptive PID controller is designed and tested using real time control experiments. The results indicate a rise time of about 150 ms to a step change in the reference force. A force variation from the reference level of about 1 N peak-to-peak, which produced a chamfer variation of 0.12 mm peak-to-peak, was achieved. The robot speed used in these experiments was 25 mm/s.The application of parameter adaptive control is examined as a solution to the problem of deburring complex contours at higher robot speeds, and to improve the robustness of the control. Simulation results are presented for an adaptive version of the Smith predictor and a form of model reference adaptive control. The results suggest that the use of adaptive control could provide significant improvements in the speed of response and robustness to sudden changes in plant dynamics, over the non-adaptive controller.