Multi-UAV Emergency Scheduling Under Demand Accumulation and Payload-Dependent Energy

Emergency delivery by multiple unmanned aerial vehicles (UAVs) requires scheduling decisions that account for demand accumulation, payload-dependent energy consumption, and the feasibility of successive sorties. To address this coupled problem, this paper proposes an energy-aware discrete black-winged kite algorithm with local search and insertion screening (EDBKA-LS-I). The model minimizes cumulative unmet demand exposure subject to payload, battery, return energy, and mission time constraints. Payload-dependent energy is calibrated from public flight records, and a sufficient speed dominance condition supports preprocessing under no-wait landing delivery. Geometric and random screening select insertion candidates for complete evaluation, while bounded local search refines UAV assignments, the sortie order, and customer sequences. Nine synthetic and public layout-derived instances were tested with twenty seeds per method under matched evaluation budgets. Three instances were additionally tested with five seeds per method under a six-second search budget. Relative to adaptive large neighborhood search with identical screening, EDBKA-LS-I reduced the mean cumulative unmet demand exposure by 7.74% and 7.93% under the respective budgets. EDBKA-LS-I also obtained lower mean exposure than variable neighborhood search on seven of nine instances under the evaluation budget. EDBKA-LS-I achieved complete service in all main comparisons. The energy model attained a mean absolute prediction error of 0.993 Wh on forty held-out delivery flights.