Research On Multi-Agent Path Planning Method In Dynamic Environment

At present,the research on multi-agent system has been successfully applied to many fields in real life,and path planning technology is one of the key issues in its research.The researchers also proposed many methods to solve multi-agent path planning.However,in real life applications,multi-agents are often in a dynamic environment.Agents have limited information on the surrounding environment.The location of dynamic obstacles is also unknown,so it is necessary to use effective algorithms to the Agent plans a valid path so that the Agent can quickly find the shortest path from the start point to the end point in a safe and untouchable situation.In addition,the collision problem between multi-agents in the path planning process is also not negligible.Aiming at the problems existing in the path planning of multi-agent in dynamic environment,this paper mainly improves the original ant colony algorithm,and combines the game theory method to solve the collision problem between multi-agents in the path planning process.The main work done is as follows:(1)Firstly,in the early stage of multi-agent path search,environment modeling should be carried out according to the surrounding environment information,and the environment modeling method should be compared.The simple grid method is used to process environmental information and establish multi-agent.The running 2D environment map lays the foundation for the implementation of the trailing path planning algorithm.(2)Using ant colony algorithm to plan global effective path for each agent,but for the shortcomings of ant colony algorithm such as slow convergence rate and easy to fall into local optimum,this paper makes the following ant colony algorithm.The first method is to introduce the reverse learning method into the original ant colony algorithm to initialize the ant position and increase the global search ability of the algorithm.Secondly,the adaptive algorithm in the particle swarm optimization algorithm is used.The inertia weighting factor adjusts the pheromone intensity Q value to make it adaptively change to avoid falling into local optimum.Finally,in order to speed up the iterative speed of the algorithm,the pheromone volatilization factor value is adaptively adjusted.The second method is: firstly,the pheromone intensity value is adaptively changed;secondly,the pheromone reduction factor is introduced based on the average of all the pheromone after the pheromone update every round,thereby speeding up the iterative speed of the algorithm;The blast operator solves the deadlock problem,thereby expanding the ant's search range and quickly finding the next path to avoid getting stuck in a deadlock.The effectiveness of the proposed method is verified by experiments in Matlab.The experimental results show that the improved ant colony algorithm obviously overcomes the shortcomings of traditional ant colony algorithm such as slow convergence rate and easy to fall into local optimum.(3)In the process of multi-agent path planning,Agent will inevitably encounter obstacles such as static and dynamic during the walking process.For the path planning problem of multi-agent in dynamic environment,Agent will have dynamic obstacle avoidance.Solve the problem of dynamic obstacle avoidance between agents.The specific method is that if there is a collision between agents,game theory is used to construct a dynamic obstacle avoidance model between multi-agents,and the problem of solving the game and the selection problem of multi-Nash equilibrium are solved by virtual action method to ensure Each Agent can quickly learn the optimal Nash balance.In the end,the algorithm can maximize the overall benefits and successfully avoid obstacles.The Matlab simulation experiment structure shows that the proposed method can reduce the path length from the starting point to the end point of the Agent path planning and improve the convergence speed.