Intelligent path planning algorithm for autonomous robot based on recurrent neural networks

Abstract

Recently, there has been increasing interest in designing autonomous mobile robots able to navigate in different types of environment and automatically avoid collisions with obstacles in their paths. In particular intelligent planning techniques have shown potential in controlling robotic fields thanks to their stability of treatment and their ability to approximate nonlinear and complex functions. In this paper, we present a path planning algorithm that allows wheeled robot to explore unknown environment. The robot would avoid collision and follow the best and shortest path towards it target. Our approach consists of developing localization algorithm for the robot in Cartesian frame, we define the position of robot for making the robot autonomous and able to predict its position regarding to the goal. Theoretical results of developed algorithm are used to generate the desirable properties of intelligent techniques and neural network has been viewed as a powerful alternative to implementation of mathematical problem. We use tow recurrent neural networks connected in series for intelligent navigation of the robot.