HI-ALWPA: A Cooperative Island-Parallel Framework for Large-Scale UAV Reconnaissance Task Allocation

Large-scale multi-UAV reconnaissance task allocation requires efficient optimization of coupled vehicle assignments and target visit orders. This study presents HI-ALWPA, a cooperative island-parallel extension of the Adaptive Lévy Wolf Pack Algorithm. The framework distributes eight operator configurations across concurrent islands and coordinates their search through asynchronous ring migration, a passive elite archive, and contribution-based population transfer. In the tested synthetic scenarios with 60–200 UAVs, HI-ALWPA reduces the mean objective by 19.07–26.60% relative to serial ALWPA under the prescribed iteration schedules. Five-seed results on ten instances per scenario show reductions of 15.61% and 24.73% in the 60- and 100-UAV cases, respectively. Ablation results support the contribution of the combined island-coordination mechanisms. In exclusive 16-vCPU measurements, HI-ALWPA achieves an iso-iteration timing ratio of 2.69 relative to serial ALWPA on the 100-UAV/200-target scenario. These results support cooperative island parallelization as an effective extension of ALWPA for the tested large-scale static UAV allocation settings.