Application And Research Of Dual Population Ant Colony Algorithm Based On Heuristic Reinforcement Learning

As one of the most important fields of scientific and technological transformation,how to improve the intelligence and work efficiency of robots have been the focus of scholars' research.Efficient path planning is a basic and critical step in the process of achieving the above goals.In solving the problem of robot path planning,ant colony algorithm has been proved to be a highly efficient algorithm.However,most improvements come from single population ant colony algorithm.When faced with more complicated environments,the performance degradation is obvious.This paper starts with the classic ant colony algorithm.Firstly,based on the prior knowledge,it establishes an effective heuristic function and improves the single population ant colony algorithm.Secondly,based on the theory of reinforcement learning,a dual population ant colony algorithm with heuristic reinforcement learning mechanism is proposed.And the advance of the algorithm is verified through the TSP problem.Finally,the dual population ant colony algorithm is applied to robots' path rules to verify the performance of the algorithm.Specific research are as follows:Firstly,An improved heuristic ant colony algorithm has been proposed in this paper to solve the TSP problem in ant colony system,which can easily be trapped in localoptimization and low convergence rate.Firstly,the threshold value of pseudo random factor is given in the early stage of iteration,so that ants can select roulette method to complete the solution construction with a high probability,and expand the range of searching for solution.At the same time,the iterative optimal ant is introduced to update the global pheromone to further increase the diversity of solution,so that the algorithm can avoid involving local extremum.Secondly,as the variation range of pseudo-random factor parameter is accelerated in the late iteration,the optimal ant is used to replace the iterative optimal ant so as to accelerate the search process to converge to the optimal solution quickly and accelerate the convergence speed.At last,the experimental results indicate that the improved algorithm can effectively skip the local optimum in the early stage,and can obviously improve the convergence speed in the later stage.Secondly,a heterogeneous dual population ant colony algorithm based on heuristic reinforcement learning is proposed.Ant colony can be divided into main population and sub-population.The main population is responsible for solution construction and pheromone update.The sub-population replaces the solution set of the main population while constructing the solution.At the beginning of the algorithm,heuristic operator is used to control the communication frequency of two populations adaptively,and the solution exchange mode is controlled by the deviation degree coefficient.At the early stage,the optimal solution of the subpopulation was allowed to replace the random solution of the main population,increasing the diversity of solutions.Meanwhile,the reinforcement learning mechanism was introduced to reward the pheromones in the optimal path of the main population adaptively,so as to increase the probability of the optimal public path being selected later.In the later stage,the optimal solution of the sub-population is controlled to replace the worst solution of the main population,the quantity of pheromones in the optimal path is strengthened,and the pheromones in the optimal path of the main population are rewarded,so as to further improve the convergence speed of the algorithm.Finally,through comparative experiments of TSP problems,it is verified that the algorithm can achieve a better balance between convergence and diversity.Finally,throughcomparative experiments of TSP problems,it is verified that the algorithm can achieve a better balance between convergence and diversity.In the end,the dual population algorithm of heuristic reinforcement learning is applied to robot path planning.First,the grid method is used in comparative experiments to verify the improvement of the algorithm;then the plug-in in the ROS system was used to apply the algorithm to the path planning of the robot turtlebot to verify the effectiveness of the algorithm;finally,a simulation environment was set up through the Gazebo simulation platform.Through turtlebot scanning,the two-dimensional matrix of environment information was obtained,and the matrix was numerically processed to get the raster map of the simulation environment,which would further verify the performance of the algorithm in the complex environment.The experimental result proves the feasibility of the algorithm.