Optimization of Multi-UAV Material Delivery Scheduling after Disaster Based on Response Levels

Abstract

Efficient relief supply delivery within the golden 72 hours after a disaster is critical to reducing casualties. For medium-to-large unmanned aerial vehicles (UAVs), pre-flight preparation—including payload mounting, center-of-gravity confirmation, flight route distribution, and link/airspace checks—is abstracted as response levels. Traditional scheduling methods often assume that UAVs are always on standby; however, maintaining the highest response level continuously is costly and unsustainable. This paper proposes a response-level–permutation coupling model for multi-UAV material delivery scheduling after disaster and designs a multi-start permutation neighborhood search (MPNS) algorithm. MPNS encodes task service order as a full permutation, uses a high-fidelity disaster simulation as the decoder, and jointly searches task permutations and active response-level control parameters. Comparative experiments against random, greedy, and genetic baselines, together with time-window sensitivity and ablation studies, demonstrate the effectiveness of MPNS in improving urgency-weighted task completion while explicitly balancing response-level operating cost.