Research On Decentralized Integral Sliding Mode Optimal Control For Reconfigurable Robot Systems Based On Adaptive Dynamic Programming

Reconfigurable robots can recombine and configure their own configurations according to different task requirements,thus showing many advantages that traditional robots do not have."Modular" and "reconfigurable" are the two basic properties for reconfigurable robot design.The main idea is to decompose a complex robot system into multiple subsystems with high portability and maintainability.Shorten the design and manufacturing cycle of the robot system effectively.In the practical application of robots,it is inevitable to encounter situations such as uncertain task environments,so that it is necessary to design appropriate control laws for the above situations.While ensuring the stability and accuracy of the robot system to complete the task,energy consumption is also a question worth considering.Optimal control is an important part of modern control theory.The core problem of its research which is for a given controlled system,and the control law is selected to optimize certain performance index of the system.For a reconfigurable robot system,in order to obtain its optimal control strategy,it is necessary to solve the Hamilton-Jacobi-Bellman equation,which is a kind of nonlinear partial differential equation,and it is difficult to obtain the optimal solution by analytical methods.Adaptive dynamic programming method is a powerful tool to solve the optimal control problem of nonlinear systems.In adaptive dynamic programming systems,neural networks are designed to approximate the performance index function and estimate the solution of Hamilton-Jacobi-Bellman equation.In order to improve the control performance of the reconfigurable robot system and reduce the energy cost of the controller,this paper combines adaptive dynamic programming method and optimal control theory to develop an optimal control system for a reconfigurable robot.Therefore,this paper forces on the following researches of reconfigurable robots:(1)Dynamics modeling of reconfigurable robot systemsFor traditional reconfigurable robots and torque sensor-based reconfigurable robot systems,the dynamic model is established and analyzed,considering the forces/torque which enforces on the joints and links and the couples force/torque among the joints.Based on local joint dynamic information,the dynamic model of the robot is described as a synthesis of interconnected subsystems,in which the dynamic characteristic is analyzed,and the dynamic model of each module is established.(2)Decentralized integral sliding mode optimal control based on adaptive dynamic programmingThe dynamic model of the reconfigurable robot in this paper has studied an optimal control algorithm based on integral sliding mode,which solves the problem of fast and stable trajectory tracking of robots and the identification of interconnected dynamic coupling terms between subsystems.This paper extends the adaptive dynamic programming method to the decentralized integral sliding mode algorithm,and combines a neural network identifier with adaptive dynamic programming to form a critic-identification structure and solves the interconnected dynamic coupling term between subsystems of optimal identification and compensation problems.Based on adaptive dynamic programming and online policy iterative algorithms,the Hamilton-Jacobi-Bellman equation is solved by a single critic network to obtain the optimal admissible control strategy.(3)Decentralized robust zero-sum optimal control in uncertain task environmentsBy adopting reconfigurable robot system dynamics model based on joint torque feedback technique,optimal control algorithm based on two-player zero-sum game is given for the robot facing uncertain task environment,and the position and speed tracking are solved.In this part,one propose a zero-sum neural-optimal decentralized control method for a reconfigurable robot system facing uncertain environment.This transforms the robust control problem of the reconfigurable robot system for the unknown environment into an optimal control problem,and based on the adaptive dynamic programming algorithm,a novel zero-sum neural-optimal control method is proposed.Based on Lyapunov theory,the closed-loop robot system guarantees asymptotic stability under the proposed decentralized control law,and finally verifies the effectiveness of the algorithm on the experimental platform.