Path Planning Based On Dstar And Artificial Neural Network For Mobile Robots In Unknown Environment

It has been deeply influenced, both industry and our life, by the emergence of robots. The technologies about intelligent robots have always been in the limelight of scientific research. The definition of robot, from scientists of our country, is that, a robot is a machine of autonomy that different from ordinary machine. It has some intelligent abilities like human or animals, such as perception, reasoning, planning, action, and collaboration. Mobile robots is such a kind of robots that could acquire the information of its environment via sensors, perform self navigation, and achive some specific tasks. Self navigation is an essential technique for mobile robots. It is also one of the representative for the degree of intelligence. Typically, navigation system should include map-learning phrase, self-positioning phrase, path planning phrase, and obstacle avoidance phrase. To some extent, there are some overlap between path planning and obstacle avoidance. Also, there are inter-depence in accuracy between map-learning and self positioning. Path planning is one of the core component of navigation system, greatly influence the performance of the robot.This study covers the following aspects:(1) In this thesis, many classical path planing methods have been scrutinized carefully by simulation experiments. The drawbacks of classical methods revealed. The origin of every methods have been researched so as to point out the key ideas, combining the classical methods and intelligent methods to form hybrid intelligent approach, to solve the problem of path planning.(2) With the research and analysis of classical and intelligent methods of path planning, a hybrid approach based on DSTAR algorithm and Artificial Neural Network has been proposed. In the proposed method, a deliberate/reactive architecture has been designed. DSTAR algorithm functions as global path planner with the ANN responsible for local path planning and obstacle avoidance.(3) Based on the proposed approach, experiments have been conducted in 3D simulation environment. Comparasion has been made between the proposed method and the traditional DSTAR. Additionally, analysis conducted in four measurements, include the length of the path computed, the amount of map information updated during path planning, the computation time, and the times of replanning, shows the feasibility and effectiveness of the proposed approach. The proposed method has the ability to tackle or avoid some problems occurred in traditional DSTAR algorithm.