Neural network controller design for a mobile robot navigation; A case study

Abstract

Mobile robot are widely applied in various aspect of human life. The main issue of this type of robot is how to navigate safely to reach the goal or finish the assigned task when applied autonomously in dynamic and uncertain environment. The application of artificial intelligence, namely neural network, can provide a "brain" for the robot to navigate safely in completing the assigned task. By applying neural network, the complexity of mobile robot control can be reduced by choosing the right model of the system, either from mathematical modeling or directly taken from the input of sensory data information. In this study, we compare the presented methods of previous researches that applies neural network to mobile robot navigation. The comparison is started by considering the right mathematical model for the robot, getting the Jacobian matrix for online training, and giving the achieved input model to the designed neural network layers in order to get the estimated position of the robot. From this literature study, it is concluded that the consideration of both kinematics and dynamics modeling of the robot will result in better performance since the exact parameters of the system are known.