Research On Neural Network-Based Sliding Mode Control For A Class Of Euler-Lagrange Systems

With the rapid advancement of science and technology,robotics has made significant progress,and robots are now capable of semi-automatic and fully automatic work,providing assistance or even replacing humans in various fields.Many mechanical systems in robots can be classified as Euler-Lagrange systems,which are nonlinear systems described by the Euler-Lagrange equation,with a wide range of applications in modern control theory and engineering.However,traditional control methods may not be suitable for Euler-Lagrange systems due to their highly nonlinear and complex dynamic characteristics.In this thesis,we focus on two typical Euler-Lagrange systems,i.e.,quadrotor and robotic manipulator system,and propose a new control strategy.Firstly,we consider a quadrotor with time-varying loads subject to unknown external perturbations.By linearizing the model,we obtain an underdriven model with external perturbations and variable loads.Then,we use a Radial Basis Function(RBF)neural network to approximate the unknown part of the system model and estimate the total airframe mass and external disturbances by adaptive techniques to compensate for the effects of load variation and unknown disturbances.The linear sliding surface and tracking error are used to establish relevant sliding equations,and the corresponding tracking controller is designed based on the neural network output signal and the adaptive signal to achieve asymptotic tracking of the reference signal under time-varying load and disturbance for the quadcopter.We prove and verify the effectiveness of the control method through the Lyapunov stability theorem and simulations.Secondly,we model the robotic manipulator with external perturbation and model uncertainty using the Euler-Lagrange method.We apply a neural network and design an adaptive sliding mode controller to compensate for the uncertainty and external perturbation of the robotic manipulator dynamics model.We then improve the controller to eliminate the system chattering caused by the sliding mode control.We demonstrate through Lyapunov stability theory analysis that the controller can stabilize the robotic manipulator system and achieve bounded tracking.The generalized Euler-Lagrange model is obtained from the dynamics equations of the robot manipulator,which further demonstrates the effectiveness of the proposed control strategy for a series of Euler-Lagrange systems with different physical backgrounds.Finally,we further improve the neural network-based sliding mode control strategy and propose a neural network-based fixed-time sliding mode control method for the Euler-Lagrange system.We solve the problem of unknown dynamics and external disturbances of the system by approximating the function with the neural network and adaptive parameter estimation while avoiding singular values in the sliding mode control.According to the proposed control strategy,the system state is guaranteed to reach the sliding mode surface in a fixed time,and the fixed-time trajectory tracking control of the Euler-Lagrange system is achieved.After stability analysis and numerical simulation,we prove and verify that the proposed neural network-based adaptive controller ensures the system tracking error converges asymptotically in fixed time.