Multi-robot task allocation for optional tasks with hidden workload: Using a model-based hyper-heuristic strategy

Abstract

Multi-robot task allocation (MRTA) is a classical problem in multi-robot systems. This paper analyzes the situations where the objective of the robots is to minimize the time cost of completing a certain proportion of tasks instead of completing all tasks, i.e., the tasks are optional. Besides, in this problem, the true workload of each task is initially hidden and can only be known after the preliminary workload is completed. As the tasks are optional, selecting a suitable combination of the tasks is quite important. The main challenge in this problem is that the robots cannot exactly select the tasks that are easy to complete because the true workload is hidden. In previous similar problems (i.e., MRTA with incomplete information), reallocation-based method is a general method. However, in this problem, if the robots exactly reallocate the tasks after collecting enough information, the sunk costs (i.e., the tasks that have been partially performed but are not selected in reallocation) limit the decrease of the time cost. Thus, we design a new hyper-heuristic strategy. In detail, a simple heuristic method that lets the robots appropriately give up some allocated tasks is combined with reallocation to design the low-level heuristic (LLH). The high-level strategy (HLS) seeks the optimal setting of LLH based on a meta-heuristic algorithm 11We respectively test particle swarm optimization (PSO) and simulated annealing (SA).. The strategies are tested based on various random instances, and the hyper-heuristic strategy can outperform the benchmark strategies in most instances. In some instances, the maximum improvement of results led by the hyper-heuristic is more than 9%.