Research On The Mobile Robot Path Planning In The Complex Enviroment

Mobile robot path planning is that a robot figures out a path from start position to goal position and avoids collision in an environment which contains some obstacles. At present, many scholars have put forward their methods and many achievements have been created. There exists many strategies to solve the problem, but this type of problem belongs to a NP-Hard problem, to find a superior algorithm for the problem becomes a study hotspot in the field. In present thesis, firstly, a new rapidly-exploring random tree algorithm for robot path planning in a known and complex environment based on grid method is proposed and its convergence is proved. The algorithm explores the space and adds a new node to a random tree of which root node is position of robot until the leaf node of the tree contains the goal node. The path composed of the edges from initial node to the goal node is what the robot will walk by. Secondly, a novel algorithm for robot rolling path planning is presented. In the algorithm, the target is mapped into the closest node inside the robot visual domain and a static, locally optimal path which robot makes a step following by is accomplished. The procedure of the algorithm is repeated on every locomotion step of robot and robot can find and follow a globally optimal path to approach the target without the risk of collisions. Finally, based on Ant Colony Optimization (ACO), this thesis presents a novel algorithm underlying the robot multi-objects path planning and dynamic obstacle avoidance in a complex and unfamiliar environment, and the algorithm optimizes the combination of all objects and can get a globally optimal navigation path. The locally optimal path between the position of robot and the present object which is gotten from the globally optimal navigation path is accomplished by local path planning with multi-ants in cooperation. Collision prediction proceeds timely after ever step of robot and local dynamic planning for obstacle avoidance is executed. Our computer experiments demonstrate that the algorithms are robust and stable.