Task allocation and robot repair strategy optimization in multi-robot system with ability decay

Multi-robot task allocation (MRTA) is an essential problem in multi-robot systems. It is worth noting that letting the robots replace humans to perform some tasks in severe environments has long been an important expectation. In severe environments, the abilities of the robots may also be influenced. Some previous studies have analyzed the influence of environments on the abilities of robots. However, they just try to present robust task allocation strategy instead of discussing the situation where the ability of a robot can be recovered after the robot is repaired. In this paper, we analyze the multi-robot task allocation problem where the abilities of robots will decay over time but the robots can select to return to the base and be repaired. There are two new challenges: (1) the size of the solution space becomes much larger because of the additional decisions about repairs; (2) the plan and expected time cost of a robot will be influenced by other robots because of repair resource preemption. The second challenge means that the time complexity of accurately calculating the expected time cost will be larger. We fully analyze this problem and present a hybrid algorithm based on collaborative discrete bee colony algorithm, simulated annealing and two heuristic modules. In experiments, the comparative analyses show that the hybrid algorithm significantly outperforms the benchmark strategies. The ablation experiments show that different modules of the hybrid algorithm respectively play the most important role in different cases.