Research On Trajectory Tracking Control Of Robots Based On Extreme Learning Machine

In recent years, intelligent control of robot systems has become the frontier of research in the field of robot control, and has attracted much attention all over the world. In the actual engineering problems, the robot systems has often the characteristics of highly nonlinear,uncertain, parametric time-varying and strong coupling, and is often affected by payload perturbations and uncertainly external disturbances. Therefore, research on the intelligent control problem for uncertain robot system has great significance in theory and practical applications. Neural networks(NNs) adaptive control has been widely used to study trajectory tracking control of robots, extreme learning machine(ELM) as a single-hidden layer feedforward neural networks(SLFNs) is supposed to be good candidates for trajectory tracking intelligent control of robots. ELM for SLFNs, which randomly chooses hidden node and its parameters, and analytically determines the output weights of SLFNs, tends to provide good generalized performance at extremely fast learning speed. ELM and NNs adaptive control theory, robust adaptive control theory and Lyapunov stability analysis theory will be introduced in this thesis to investigate the control problem of robot systems and propose ELM adaptive neural control methods. The main contributions of this thesis are as follows:(1) The basic structural characteristics of ELM is analyzed and its learning algorithm is further studied, and the ELM network different from traditional SLFNs is presented. Finally,the ELM network is applied to online identification of a nonlinear dynamical systems. Simulation result show that ELM network have good identifying capacity.(2) Based on ELM, an adaptive neural control method for the uncertain rigid arm robot system is presented. Using the Lyapunov syntnesis approach, The designed ELM controller,which guarantee the stability of the overall closed-loop control system, can approximate the model uncertainy of systems by adaptively tuning the output weight. The proposed adaptive neural controller is finally applied to control a planar manipulators with two degrees of freedom and compared with existing radial basis function neural networks(RBFNNs) method.Experiment result show that ELM neural controller demonstrates the effectiveness of the proposed control method and has good tracking performance.(3) Based on ELM, two adaptive neural control methods for uncertain rigid arm robot are presented in this paper, and achieve adaptive tracking control of rigid manipulators in task space. Within these adaptive control methods, ELM is employed to approximation the plant’s unknown nonlinear function and robust control term is used to compensate for approximation error. Parameter adaptive laws and robust control term of ELM controllers are derived based on Lyapunov stability analysis so that global stability and asymptotic convergence to zero of tracking errors can be guaranteed. Futhermore, two adaptive controllers do not depend on anyparameter initialization conditions and relax the repuirement of bounding parameter values.Finally, the proposed adaptive ELM control methods are applied to a tracking control instance for two-link rigid manipulators and compared with existing RBFNNs control methods. Simulation result show that ELM controllers have good tracking performance and demonstrate the effectiveness of the proposed control algrorithms.(4) Based on ELM, an robust adaptive neural control method for a class of uncertain continuous-times multiple input multiple output(MIMO) affine nonlinear dynamic systems is presented. In the proposed adaptive control, the assumption on nonsingulatity of ELM approximation for unknown control coefficient matrice is elminated, meanwhile, output weights of ELM, unknown upper bound values of approximation errors and external disturbances can be online estimated through parameter adaptive laws using Lyapunov stability analysis so that semi-global uniform ultimare boundedness of all signals in the closed-loop system can be guaranteed. Compared with existing method such as RBFNNs adaptive controller under the same condition, simulation and experimental results are provided to verfy the effectiveness of the proposed robust adaptive ELM control method.