Using an ant algorithm to find a strategy for the behavior of a group of mobile robots on a work field with obstacles

Abstract

A more complex solution to the problem of the distribution of tasks in a group of mobile robots, in the presence of obstacles in the workspace, is considered. The work is a continuation of a cycle of research in which the basic algorithm for solving the tasks was one of the types of ant algorithm — the multicolonial ant system method in combination with the trajectory planning algorithm implemented using the principle of dynamic programming. The task statement, the workspace model, the goals of the robots functioning and the parameters characterizing their work have been adjusted. The choice of free parameters of the ant algorithm for performing multi-criteria optimization and tuning of the solution is carried out: the number of iterations, the number of intercolonial groups of ants, the weight of the concentration of the pheromone of arcs, the weight of the heuristic attractiveness of arcs and the pheromone evaporation coefficient. The results of computational experiments conducted in the presence of static (known in advance) and dynamic (other robots) obstacles in the workspace are presented. The proposed algorithm was tested using the example of a group consisting of three robots performing 10 tasks. As shown in the results of computational experiments, robot trajectories are built on a subset of free cells of the workspace and do not intersect cells with obstacles. At the same time, the configuration of the work field affects not only the actual routes of robots, but also the redistribution of tasks between them, and the number of robots involved. Additionally, a series of computational experiments with different combinations of values of free parameters was carried out to determine the optimal ratios and implement a more efficient ant algorithm. Optimization was carried out by a single adjustment method, which allowed us to find the required values of free parameters. It is shown that the adjustment of the parameters made it possible to reduce the relative error in the synthesis of the optimal route of movement of a group of robots by 3–6 %.