An improved genetic algorithm for multi-robot task assignment problem

Abstract

The multi-robot task assignment problem is always a popular topic in the field of robotics, which is used in numerous applications. The main purpose of multi-robot task assignment is to optimally give a series of tasks to robots in the system to optimize the performance of the system and automate the robot team. The current algorithms often focus on simple tasks, which can be straightforwardly executed by a robot. In this dissertation, an improved genetic algorithm based on GAHIR is proposed to solve more complex MRTA problems, in which the complex tasks need to be completed by more than one robot. This dissertation firstly establishes the mathematical model of the MRTA problem, then proposes three improvement strategies for overcoming the shortcomings of premature convergence, high computational cost and low efficiency of the original GAHIR algorithm. Finally, multiple simulation experiments are designed on MATLAB to evaluate the performance of the improved genetic algorithm. The experimental results show that the improvement strategies can effectively maintain the diversity of the population, enhance the ability to jump out of the local optimum, and prevent the algorithm from premature convergence. In various experimental scenarios, the performance of the solutions generated by the improved genetic algorithm is significantly higher than that of other advanced assignment algorithms.