Research On Decentralized Control And Active Fault Tolerant Control For Reconfigurable Manipulator

Different from the traditional manipulator, according to the task needs,reconfigurable manipulator can be combined into different configurations to performdifferent tasks. Beacuse of the characteristics of the modular, reconfigurablemanipulator has the strong adaptability, low cost, good fault tolerance, high stability.Nowadays, reconfigurable manipulator has been widely applied in the field ofdangerous condition working, military, space exploration, industrial, medical,entertainment, etc, and instead of the human, it can complete many works that can notbe completed by human. Therefore, research on reconfigurable manipulator hasimportant theoretical value and practical significance.Based on the modular characteristics, when reconfigurable manipulator motionsin free space, decentralized trajectory tracking control of the each joint is needed. First,the complex dynamics system of reconfigurable manipulator is decomposed into jointsubsystems, then design controller for the each subsystem respectively, and localinformation of its own joint is only needed for subsystem controller. But how torestrain the coupling effect between subsystems has become the key problem ofdecentralized control; In the stability proof of the control system by using theLyapunov stability theory, a general algorithm of constructing Lyapunov function islack. The backstepping control method solves the problem that Lyapunov function isdifficult to construct, but how to solve the "calculation explosion " problem has been abottleneck restricting the development of backstepping method; The big jump of thedesired trajectory may cause the velocity jump (it needs great acceleration to achievevelocity jump, the corresponding driving force or torque will be great, this is oftenbeyond the maximum of the motor), how to solve this problem decides that whetherreconfigurable manipulator can smooth tracking desired trajectory in the practicalapplication. When reconfigurable manipulator replaces human to completeteleoperation and dangerous tasks such as space mission, maintenance of nuclearfacilities, the sensor and actuator faults are inevitable.The faults will cause thatreconfigurable manipulator can not complete the task, and even lead to seriousconsequences. In order to improve the reliability and the ability of self repairing of the manipulator, fault identification and fault tolerant control technology has become anurgent research topic; When reconfigurable manipulator implements precisionassembly, capture, handling and other tasks, in addition to control the joint angle andthe end position, as well as the contact force with the environmental. Because eachjoint of the reconfigurable manipulator is lack of the torque sensor, which is bound toaffect the effect of feedback control. In addition, the traditional robust control mustaccurately realize controller, which lead to the poor robustness of the controller itself,so how to improve the robustness of the controller itself has become a key factoraffecting the practical application of manipulator.This thesis mainly studies on the above problems. The kinematics and dynamicmodel of reconfigurable manipulator, inverse kinematics solution of reconfigurablemanipulator based on improved PSO, backstepping decentralized control based on theESO and DSC, adaptive backstepping fast terminal fuzzy sliding mode control basedon the bio inspired strategies, active fault tolerant control for the actuator fault ofreconfigurable manipulator based on iterative fault tracking observer, active faulttolerant control for reconfigurable manipulator with multiple concurrent faults, doubleclosed loop decentralized force control for reconfigurable manipulator based onnonlinear joint torque observer and non-fragile decentralized robust force/positioncontrol based on LMI were researched. This thesis is organized as follows,in whichthe main research works are included.1、The research significance has been presented first, then overviews the researchstatus of reconfigurable manipulator at home and abroad, as well as the key problemsof the research.2、Kinematics product formula of reconfigurable manipulator based on the spinortheory is deduced; The inverse kinematics solving process based on the improvedPSO is detaildly introduced, and the verification results are given; Introducing spiralmotion, the dynamic model of reconfigurable manipulator based on the Lagrangeequation is established.3、Based on the modular characteristics, the each joint of reconfigurablemanipulator is regarded as a subsystem.Aiming at the decentralized trajectory trackingcontrol of reconfigurable manipulator, decentralized adaptive fuzzy controller basedon ESO is designed, and coupling interconnection is estimated by using ESO.On thisbasis, the backstepping decentralized controller based on the ESO and DSC isdesigned, thus solve the problems that the Lyapunov function is difficult to construct and “calculation expansion” in the backstepping control; In order to solve velocityjump problem caused by the state jump of the desired trajectory, the adaptivebackstepping fast terminal fuzzy sliding mode controller based on the bio inspiredstrategies is designed.4、For the actuator fault of the each joint subsystem, iterative thought isintroduced into the fault identification. In order to real-timely observe fault,the faulttracking observer is designed, and then actuator active fault tolerant controller of thereconfigurable manipulator based on the iterative learning fault tracking observer isdesigned.The uncertainty and coupling interconnection between subsystems areestimated by using fuzzy system, and the estimation error of the fuzzy system isadaptively compensated.When the actuator fault occured, reconfigurable manipulatorcan still track the desired trajectory; For reconfigurable manipulator with multipleconcurrent faults, fault reconstruction and active fault tolerant controller based on thesliding mode observer are presented. First, a new state that make sensor faulttransform into actuator fault is introduced. And then a sliding mode observer isdesigned, in which the interconnections between subsystems and uncertainties areapproximated by the fuzzy neural network with the adjustable width and center, so asto realize the reconstruction of different sensor and actuator faults. Finally,the faultsensor output is replaced by sliding mode observer output, so fault tolerant control canbe timely implemented when faults occur.5、 In the process of completing the capture, handling and other tasks,reconfigurable manipulator will contact with the environmental (be called constrainedreconfigurable manipulator), it is need to control position and force at the same time.Considering the modular characteristics of reconfigurable manipulator and existingjoint module without torque sensor, double closed loop decentralized adaptive forcecontrol method based on the nonlinear joints torque observer is proposed.The fuzzysystem is used to estimate coupled interconnections between subsystems, end contactforce is mapped to each joint subsystem through the Jacobi matrix so as to designsubsystem controller.The double closed loop regulation of control input forreconfigurable manipulator subsystems is constituted by means of the contact forceerror and the error between the joint torque and its observer, so the contact force iscontrolled and improve the convergence speed and tracking accuracy; In order toimprove the robustness of the controller itself, based on the Lyapunov stability theory and Linear Matrix Inequality (LMI) method, the non-fragile robust decentralizedforce/position controller is designed.When the controller parameters change within acertain range, the system can still ensure the stability and satisfyH performanceindex.Finally, the main research results of this thesis and the perspective of the futureresearch are given at the end of this thesis.