Multi-UAV Task Allocation Based on Grid-Based Particle Swarm and Genetic Hybrid Algorithm

Abstract

To address the uneven distribution of the Pareto front and insufficient convergence in multi-UAV task allocation, this paper proposes GrEAPSO, an improved algorithm that hybridizes Particle Swarm Optimization (PSO) with Genetic Algorithm (GA). GrEAPSO balances exploitation and exploration through grid partitioning, adopts a dual-encoding scheme coupled with crossover and mutation to enhance population diversity, and employs a grid-based environmental selection mechanism to improve the uniformity of the Pareto set. After initialization, the algorithm iteratively performs a PSO-based local search, genetic crossover and mutation, and grid-based environmental selection. The offspring and parent populations are then merged, and the archive set is updated accordingly. Across three military UAV task-allocation scenarios (small, medium, and large), GrEAPSO is benchmarked against MOPSO, NSGA-II/III, MOEA/D-DE, RVEA, IBEA, MOMVO, and MaOGOA. All experiments use a population size of 100. Its reference point is undominated and dominates some competitors, with median gains of 55.78% in hypervolume and 8.11% in spacing. Finally, the sensitive analysis further indicates that dividing the objective space into 15–20 grids offers the best trade-off between search breadth and solution distribution.