Research On Muti-objective Path Planning Method Based On Intelligent Optimization Algorithm

In recent years,with the rapid development of different kinds of intelligent mobile robots,research issues related to mobile robots have gradually become the central issue.When using mobile robots to solve the practical problems,there will often encounter many different situations,such as search task,detection task,transportation task and so on.There are many traditional path planning methods for mobile robots,such as graph search algorithm,artificial potential field optimization algorithm,etc.The swarm intelligence optimization algorithm has better effect in solving multiple optimization problems because of its parallel search and insensitive to mathematical relations.However,when solving multi-objective point planning problems,the completeness of such algorithms is limited by the degree of discretization for the map space,and this kind of algorithm many not always have a good performance in high-dimensional state space.For the situation of the limitations of traditional path planning methods for mobile robots when solving multi-objective point path planning task,this paper attempts to use an improved ant colony algorithm which based on reinforcement learning to solve the problem of multi-objective point planning for mobile robots.Firstly,for the situation of the large amount of pheromones accumulated on local optimal path nodes in traditional ant colony algorithm,this paper adds population density as an index to measure population diversity,in order to avoid most of the ant individuals gathered in the local optimal solution and ensure that the population can effectively "explore" the space at the same time.Secondly,in the iterative process of the algorithm,the strategy of dynamic division of solution space is added.According to the distribution of pheromones,the exploration space of ant colony is adjusted adaptively,so as to improve the exploration of the better solution space and the search efficiency of the algorithm.Then,the state information of ant colony is used to build reinforcement learning model,and corresponding environment state space,action state space,reward function,action selection strategy are established.In the iterative process of ant colony algorithm,strengthen the learning model to find the balance between "exploration" and "utilization" in the process of ant colony algorithm path planning,adjust the update strategy of pheromone,and then improve the search ability of the algorithm to get a more reasonable solution.Finally,the algorithm is applied to the comparative experiment of multi-objective point path planning,and the simulation experiments and results of different map environments and different target points are compared.The standard test function in CEC2017 is used to test the algorithm proposed in this paper,and compared with other three ant based algorithms.The experimental results show that the algorithm in this paper has better optimization ability and stability in solving continuous optimization problems,and can effectively overcome the shortcomings of classical ant colony algorithm that is easy to fall into local optimization.In the comparative simulation experiment of multi-objective point path planning,three algorithms are tested in four different map environments.The experimental results show that the algorithm proposed in this paper has the following characteristics: good real-time performance,high stability and high planning efficiency.