DESIGN AND DEVELOPMENT OF TRACTION TRACKING CONTROL FOR MOBILE ROBOTS BASED ON NEURAL NETWORKS SUBJECT TO UNCERTAINTY PARAMETERS AND DISTURBANCES

Abstract

The paper presents the design and development of mobile robot trajectory tracking controllers based on neural networks subject to uncertain parameters
and disturbances. The algorithm has a control structure that integrates a backstepping controller and a neural controller for mobile robots. The mobile robot
system has been modelled, including the robot's kinematic, dynamic, and actuator models. The dynamic model is calculated using the Lagrangian method for
nonlinear systems. When the dynamic parameters of the robot are unknown or not precisely determined, the neural network is chosen as a two-hidden layer
neural network with the assumption of linear activation functions for the output layer. The output of each neuron in the hidden layer can be calculated based
on the network input, the weight of the first layer and the activation function of the hidden layer. The calculation ensures that the error signal is zero and that
the Lyapunov method proves the stability. Different mobile robot controllers have been simulated using Matlab/Simulink. As a result, the trajectory tracking of
mobile robots by the controller using a neural network (NN) gives better trajectory tracking results than the PD-Backstepping controller in the presence of
uncertain parameters and model disturbances.