Vibration Control Of A Flexible Robotic Manipulator Based On Adaptive Neural Network

With the development and requirement of times,the control of robotic systems has the tendency of fast speed,heavy load,high accuracy and low energy consumption.Therefore,the study of the flexible robotic manipulator with light weight,high flexibility and low energy consumption gradually attracts many researchers' attention.Based on the characteristic of flexible structure,the flexible robotic manipulator system is a distributed parameter system and the system is complex so that model of the flexible robotic manipulator is commonly described by partial differential equations(PDEs).In order to achieve control objective of tracking the desired reference with guaranteeing a good performance of vibration control simultaneously,the controller design based on PDE model is complex and hard,especially,it is more difficult when the flexible robotic manipulator is subject to system uncertainty.This paper focuses on the problem of vibration control of the flexible robotic manipulator with system uncertainty and nonlinearity.Based on the good capability of approximating the nonlinearities and uncertainties,the neural network is introduced to the control of the flexible robotic manipulator to solve the problems of uncertainties and nonlinearities.As the limitations of PDE model,the neural network control cannot be well applied to the vibration control of the flexible robotic manipulator.Therefore,the PDE model of the flexible robotic manipulator is needed to be converted to an ordinary differential equation(ODE)model so that the problem of conducting environment of neural network is well solved.And the ODE model of the system is constructed based on the lumped parameter method and assumed model method,etc.On the basis of the ODE model,the paper emphasizes on designing an adaptive neural network control method to deal with vibration control problems of a single-link flexible robotic manipulator with input deadzone and system uncertainty.The ODE model of the flexible robotic manipulator extended the control methods of vibration.And as the recently popular control method,reinforcement learning is creatively introduced to the design of vibration control in this paper.In the control design,based on reinforcement learning,its control object is to control the vibration of the flexible robotic manipulator only with system uncertainty,and the nonlinearities are not taken into consideration.For design of reinforcement learning controller,the algorithm of actor-critic structure is utilized.The critic part is utilized to judge whether the applied control is good or bad,and actor part is utilized to correct the control input based on the judgement of critic part to achieve the optimal control performance.To sum up,in this paper,the neural network technique is cleverly applied to the design of vibration control of the flexible robotic manipulator with adaptive laws designed to update the neural network weights.The vibration control based on adaptive neural network of a single-link flexible robotic manipulator with input deadzone and system uncertainty is firstly studied,and then based on this study,reinforcement learning algorithm is also introduced to the vibration control of the flexible robotic manipulator to conduct the further research such that the research can provide a good idea for the following vibration control methods of the flexible robotic manipulator.