An Adaptive Feedforward Optimization Mechanism for Improving Industrial Robots Trajectory Tracking Performance through Iterative Learning

This paper presents an adaptive feedforward optimization method based on iterative learning control mechanism, improving trajectory tracking accuracy from trial to trail. Conventional feedforward control researches focus on the precise prediction and compensation of torques, either by means of dynamic modelling and parameter identification or by data-driven approaches. However, there exists unpredictable time-delay for torques to work. The proposed method takes the previously less-concerned velocity feedforward part of the servo control system into consideration, and a self-adaptive fine-tuning mechanism is designed and established. The error measurement and feedforward optimization processes are integrated into an iterative learning framework to obtain efficiency and precision. The effectiveness of proposed method is validated by physical experiments on an EFORT ER15 industrial robot, and an averagely 90.71% error reduction is witnessed.