Research On High-speed And High-precision Control Of Industrial Robots

With the expanding field of industrial robot applications and the rapid development ofmodern industry, the performance of industrial robot has become increasingly demanding tofurther improve production efficiency and product quality. Therefore, high-speed,high-precision, intelligent and modular industrial robots become the main trends in thedevelopment of industrial robots. The industrial robot is a highly nonlinear, highly coupled,multi-input-multi-output complex system. In the case of high-speed movement, the nonlineardynamic property of the robot is quite significant, and a variety of complex uncertaintyfactors also seriously affect the tracking performance. Therefore, research on high-speed andhigh-precision control of industrial robots has important theoretical significance and practicalvalue. In this thesis, some key problems relate to the singular avoidance in work space,dynamic parameters identification and advanced robust control for6-DOF industrial robotsare systematically analyzed and studied. The main contributions are as follows:For the singularity problem of6-DOF industrial robot designed in our lab, the kinematicmodel is established to analyze singularities, and a "singularity separation plus exponentialdamped reciprocal" method is proposed to guarantee the continuity and smoothness of thehigh speed movement of the robot, it effectively reduce the influence on the accuracy of theend-effector at or in the vicinity of a singular configuration. In application, the singularityfactor is separated from the inverse Jacobian matrix, and modified with the exponentialdamped reciprocal, so the computational amount is less. Simulation results show theeffectiveness and feasibility of the method.The dynamic parameter identification of robots is the basis for the design of advancedmodel-based control system, and tracking control performances of robots depend directly onthe model accuracy. First, the dynamic model is studied to obtain the minimum inertialparameters and observation matrix of industrial robot. Second, in order to guarantee theparameters estimation accuracy and robustness of the Least Squares method, the excitationtrajectory optimization method based on artificial immune clonal selection algorithm isproposed. Simulations results validate the accuracy of the identification results.The robustness of finite-time control system is studied, and a robust H∞finite-time stability concept and stability theorem are proposed. Moreover, For a class of commonnonlinear system with uncertainty, a robust H∞finite-time stability controller is proposed toobtain finite time stability and disturbance attenuation in the L2gain sense, without solvingthe nonlinear Hamilton-Jacobi-Isaacs inequality or Riccati equation. In practice, the industrialrobot system is also affected by the various complex uncertain factors, so the robust H∞finite-time controller for industrial robot is designed, on one hand to guarantee the finite timeconvergence of tracking error, and improve the response speed and tracking accuracy; on theother hand to ensure L2-gain disturbance rejection performance for uncertainties. Simulationresults show that the proposed controller is strongly robust to the time-varying and abruptdisturbance.The robust finite-time control of robot is designed using backstepping method andterminal sliding mode. A finite-time tracking controller based on backstepping method isstudied to improve the robustness of the finite time stability system by a variable structure.Combining the nonlinear mapping ability of RBF neural network, a robust adaptive neuralnetwork finite-time controller for industrial robots is proposed to overcome the shortcomingof knowing upper bound of the uncertainty for finite-time control system. Using terminalsliding mode method, the modified neural network finite-time controller is designed tosimultaneously guarantee the finite time convergence and eliminate the chattering. The finitetime stability of closed loop system is proved, simulation results verify the effectiveness andfeasibility of the proposed controllers.When the robot dynamic model is completely unknown, the time-delay information isused to estimate the nonlinear dynamic and external disturbance. the robust H∞control basedon time-delayed estimation for industrial robot is proposed to further improve the robustnessto estimation error. Combining the advantages of finite-time control, the finite-time controlbased on time-delayed estimation is proposed to improve response speed and trackingaccuracy. Simulation results verify the effectiveness and feasibility of the two algorithms.Finally, the comparative analysis of the robust H∞finite-time control, finite-time control basedon backstepping, robust adaptive neural network finite-time control and neural networkfinite-time control based on terminal sliding mode are presented. Advantages anddisadvantages, similarities and differences of these control methods applied to high-speed and high-precision tracking control for industrial robot are analyzed respectively, especially inrobustness, transient performance and difficulties in realization, in order to facilitate theselection.For the industrial robot with kinematic and dynamic uncertainties, the fuzzy adaptivesliding mode controller is proposed. The trajectory planned in work space, is driven throughthe movement of each joint of robot, so the end-effector trajectory tracking accuracy is alsoaffected by the robot geometric parameters uncertainties. The proposed controller is to makefull use of linguistic rules information and learning approximation capability of fuzzy logicsystem, and adaptively compensate for the kinematic and dynamic uncertainties, to improvethe end-effector tracking accuracy, without estimating the kinematic and dynamic parametersof robot on line.The artificial structure in the open multi-controller robot control system is proposed tofacilitate parameter identification experiments, and automatically select the best controlleraccording to different tasks and working environment.Finally, in order to verify the effectiveness and easy engineering implementation ofparameter identification method and these control methods, a robot experimental platform isdesigned to carry out experimental studies. The experiment is divied into two parts: the fistpart is the identification experiment of the robot to obtain the accurate dynamic model; thesecond part is the trajectory tracking control experiment, the experimental results show thatthe proposed controllers can effectively improve the transient characteristics, trackingaccuracy and robustness of robot system.