Task Distribution Module for a Team of Robots Based on Genetic Algorithms: Synthesis Methodology and Testing

Abstract

One of the variants of using genetic algorithms for solving the problem of task distribution in teams of mobile robots performing single-syllable operations in a single workspace is considered. There are known solutions to this problem, implemented on the basis of market economy algorithms, neural networks, dynamic programming, potential fields, etc. All these methods assume that the number of tasks to be solved corresponds to the number of robots in a group, and the selection criterion is reduced to finding a minimax solution in accordance with the selected criteria (efficiency, energy or time costs). The algorithm proposed in the article allows solving a multicriteria optimization task taking into account time costs (time spent by a robot on moving from the initial position to the place of the task) and the integral criterion of the effectiveness of robots, taking into account energy costs, functionality of each agent of the team, etc. The specificity of the working space model is considered, the goals of the functioning of robots are formalized, and the method of defining the set of tasks solved by a team is defined. A description of the parameters that determine the effectiveness of the solution using genetic algorithms is given, a form of coding the solution in the form of a chromosome is proposed, and specific fitness functions are composed. Taking into account the class of the problem to be solved, a genetic algorithm was chosen with a sorting by domination and elite selection with integer restrictions. When these are triggered, the mutation operation is excluded from the general algorithm.