Research On Path Planning Algorithm Based On Q-IGA And Fuzzy Neural Network

Since the 1960 s and 1970 s,with the development of computer science,control theory,sensor technology,computer network technology and artificial intelligence,mobile robots developed rapidly and have been widely used to accomplish onerous and difficult tasks in industry.In order to adapt to the actual working environment and complete tasks effectively,robots' abilities of navigation,dynamic obstacle avoidance and path planning need to be continuously optimized.The motion environment of mobile robots is complex and changeable,which makes research on path planning become the focus of mobile robot motion problems.Compared with the traditional algorithms,the intelligent bionic algorithms have the advantages of strong flexibility and high computational efficiency so that it can overcome the instability and low computational efficiency of the traditional algorithms.Therefore,this paper studies the path planning based on genetic algorithm and neural network.In path planning,in order to obtain an optimal path suitable for actual situation and overcome the inherent shortcomings of genetic algorithm,such as easy converge to local optimal and high complexity,a Q-IGA-based path planning algorithm with dynamically fitted Bezier curve is proposed firstly in this paper.The proposed algorithm replaces static fitting method of directly using Bezier curve with simultaneously searching path and control points of Bezier curve.What's more,an additional judgment criterion based on Q value is added into selection operator,which can eliminate the solution with high similarity and enhance the diversity of the population.At the same time,the proposed method optimizes the fitness function by taking robot volume and turning angle into consideration,so that the selected path is not only short but also a reasonable path which keeps a safe distance from the obstacles.Simulation results show that the paths produced by Q-IGA algorithm are more reasonable than those produced by improved artificial potential field algorithm and hybrid genetic algorithm.As it can reduce the search time and the energy consumption of the robot,the proposed method is more suitable for practical industrial applications.In addition,compared with Dijkstra algorithm,A* algorithm can guide the path planning by introducing heuristic function.The basic heuristic function can't estimate the forward distance accurately because of obstacle occlusion,which makes the search efficiency low.Therefore,this paper proposes a computing method of heuristic function by fuzzy neural network.Firstly,the particle swarm optimization algorithm is used to optimize the parameters of fuzzy neural network.Then the cost of arriving at the target point is predicted by trained fuzzy neural network.Finally,the cost is added to the cost function of A* algorithm.This method can determine a more accurate search direction by estimating the influence of obstacles on the path within a certain range.In addition,it can overcome the problem that the classical heuristic function is inaccurate when the obstacles are complicated.Simulation results show that compared with Dijkstra algorithm and greedy algorithm,the proposed algorithm greatly reduces the number of traversed nodes in the search process and makes the direction search of the path more explicit.