Cooperative Task Allocation for Multi-Robot Systems Based on Multi-Objective Ant Colony System

This paper proposes a novel multi-objective ant colony system (MOACS) approach to solve the cooperative task allocation problem of multi-robot systems. The task allocation problem is formulated as a multi-objective multiple traveling salesman problem (MTSP). The objectives are to minimize the total and maximum cost of the robotic vehicles so that the workload of each vehicle could be balanced. The time cost matrices of the salesmen are different and asymmetric due to the different flight speeds of vehicles and executing time of tasks. Based on the single-objective ant colony system (ACS), a novel solution construction method and a novel pheromone update rule are proposed. At each step in the solution construction phase, the ant with minimum cost has the biggest chance to add an unassigned task to balance the workload of each vehicle, while the ant with maximum cost also has a bigger chance than any other ants to add an unassigned task to find better Pareto front. The minimum value of the pheromone is limited in the pheromone update phase, which is helpful in avoiding fast convergence and local optima. Extensive simulation results suggest that the proposed MOACS has better performance and effectiveness than the existing non-dominated sorting genetic algorithm II (NSGA-II) and multi-objective particle swarm optimization (MOPSO). Hardware-in-the-loop experiments on multiple unmanned aerial vehicles (UAVs) also show that compared with NSGA-II and MOPSO, the maximum and total flight distance of the UAVs with the proposed MOACS are decreased by up to 28.46% and 26.34%, respectively, while the maximum and total time used to finish all tasks are decreased by up to 23.86% and 17.94%.