Research On Robust Control Of Robot Manipulators Tracking Under Uncertainties

Robot is not only a high complicated time-varied, strong-coupled, nonlinear systems,but also subjected to various kinds of uncertainties, such as measurement errors, frictions,varying load, random disturbances, unmodeled dynamics, and so on. During the robottrajectory tracking, some pivotal problems, which restrict the development andapplication of robot, such as the solution of inverse kinematics and the failure of robotcomponents. It is significance in theory and practicability to the realization of high speed,high precision and great capability robots that modeling and solving the above problems.Researched on the uncertain parameters and structures of robotic model in thetrajectory tracking some effective control strategies are presented. The main contents areoutlined below:(1) For the existence and speedily convergence of inverse kinematics, thisdissertation firstly presents the method based on fuzzy-neural network considering theircharacteristics such as parallel computing, memory, association, adaptive learning etc.And this problem is solved perfectly by three controllers designed on this principle.(2) This dissertation considers adequately the uncertainties in robot trajectorytracking, and studied them deeply. Two robust-adaptive control strategies are designed.By series of transformations we transform the robot system into chained form, and designa robust adaptive controller based on passivity according to the relation between thepassivity and the asymptotic stability. The decentralized robust adaptive control strategyis improved by introducing a cube term of general error into the control of it, whichdoesn't depend on accurate dynamical model. The proposed controllers overcome thedefects of it based on the accurate model and make the tracking error attain to zerorapidly.(3) Aiming at the parameters uncertainties and external disturbances existed widelyin robot systems, we present two kinds of control strategies, which combine thebackstepping method with the control performance of H∞ control theory. It not onlyimproves the robustness of the system, but also avoids the complex calculations of HJIinequation of H∞ method in nonlinear systems that reducing the influence to systemcharacteristics by choosing the appropriate external disturbances attenuation parameterγ .The solution of nonlinear H∞ algorithm of controller is simplified greatly.(4) A robust adaptive neural network sliding model control strategy is present bycombining variable structure theory with nonlinear mapping ability of neural network.And this strategy not only overcomes the bound requirement of uncertainties present bygeneral sliding-mode controller but also guarantees the asymptotic stability. A velocityobserver approach based on fuzzy neural network control and H∞ control is proposedfor the robot manipulators, which only possess accurately position's measurement. Itrealizes robust speed observation of uncertain robot systems and reconstructs the speedsignal validly. That not only guarantees the stability of system, but also restrains theinfluences of external disturbances.(5) A compound fault tolerance control strategy is presented. While an actuator failed,it has to be locked and then robot system's dynamic model becomes a structure uncertainsystem. A robust adaptive iterative learning control strategy based on passivity is designedto attenuate structure uncertainties. Locking the joint directly may reduce the robot'sworkplace so that some operation missions will not be completed, a variable structurecontroller is presented to drive the failed joints. The failed joints will get angles needed inpractice by the dynamic coupling effect between the actuated joint and the failed jointbefore it is locked. The control strategy presented above can not only complete thescheduled missions but also keep a highly tracking accuracy.All the strategies suggested in this dissertation are proved strictly in theories;theirpracticality and validity are verified in simulation. In addition, the strategies proposed inthis dissertation not only can be used in uncertain robot control but also be applied inother analogous nonlinear mechanical systems.