Improved Ant Colony Algorithms And Research On The Problem Of Path Planning For Mobile Robot

The path planning for mobile robot is one of the cores in the robotics research. It is always a focus topic for the researchers all over the world. However, traditional methods of path planning have their own drawbacks respectively. To find a superior algorithm for the problem becomes a study hotspot in the field. In view of the advantages of the ACO in solving combinatorial optimization problems and its own characteristics, comply with the intellective and bionics tendency of the field’s development, new ACO is designed and successfully applied in solving the problem of path planning for mobile robot with the expected results. The main research work is as follows:Firstly, the grid method is used to divide the environmental space. The environment map with the grid method can avoid complex calculations in dealing with obstacles boundary, but easy to fall into the trap environment too. The ants will be in a "dead" state and result in stagnant algorithm. Therefore, a new coding method is proposed on the basis of effective vertexes of all the obstacles on the grid map to avoid that ants fall into the trap environment.Secondly, ant colony algorithm has shortcomings and deficiencies in certain areas, so the following improvements are made:(1) In the early stages of ant colony algorithm, the pheromone is equal distribution, and lead to long search time. Therefore, pheromone gain is added for the initial distribution of pheromone.(2) In order to improve the structural efficiency of the initial feasible solutions, and give full play to ants’ ability to route, a strategy of ant opposite parallel search from literature [1] is referred. On the basis of it, a new identification strategy is proposed to determine when ants meet. New identification strategy ensures that all feasible paths can be searched and make up for the shortage of losing a number of possible paths when using original identification method.(3) Pheromone mutual guidance strategy is also proposed and the route choice probability formula is redesigned in order that ants make a response to the optimal path more quickly.(4) At the same time of improving the global search capability of ant colony algorithm and avoiding falling into local optimum, the search time is also increasing. Therefore, a dynamic and adaptive adjust pheromone strategy is designed as a compromise between the two contradictions. Thirdly, several local collision-free strategies are used after the environment detection and collision prediction based on rolling windows in order that the robot reaches the goal safely.Finally, to validate the improved algorithm is effective and feasible, lots of emulation studies in2D barricade environment were realized by MATLAB. Simulation results show that the proposed algorithm can quickly complete the task of path planning in complex space and has strong robustness, strong ability of global optimization. Meanwhile, algorithm complexity is moderate and higher efficiency.