Improved Backstepping Controller for Rigid-Flexible System using Input Shaping Reference Model Matching and Neural Network

Abstract

Backstepping is a good control technique for rigid-flexible system, but it has some limitations that hard to use in real control applications. First, a desired trajectory should be a continuous function and not change rapidly. Second, math model used in design process should be exactly precise. This paper presents a novel backstepping control method for rigid flexible system. The proposed approach consists of 4 techniques. First, main controller is based on backstepping technique. Second, a reference model matching is used to convert the desired trajectory to ensure that it is suitable for the backstepping controller. Third, input shaping method is used to improve a speed performance of reference model matching. Finally, uncertainty and disturbance in the system is compensated by using Neural Network. Simulation results on a one link Flexible Joint Robot manipulator (FJR) show that the proposed approach can improve performance of traditional backstepping method which can be described as follows. First of all, it can be applied when the desired trajectory is discontinuous. Moreover, the response of the system using the proposed control approach is better than the control system using normal backstepping control approach. Finally, the proposed approach can also be used to control uncertain system.