A comparative study of path planning algorithms for low-altitude emergency medical UAVs considering task priorities

In urban emergency medical scenarios, unmanned aerial vehicle (UAV) path planning faces severe challenges from superimposed uncertain factors: dynamic emergency task insertions, sudden no-fly zones (NFZs), and day-to-night environmental transitions. While existing studies frequently address these issues in isolation, there remains a critical lack of systematic benchmarking frameworks that incorporate these stress factors simultaneously. Based on Guangzhou geographic data, this study constructs a medical network with 6 blood centers and 12 hospitals, designing six composite stress-testing scenarios. Seven algorithms spanning a three-tier capability gradient (exact MILP, metaheuristics MSHOA/ACO/GA/PSO, and learning-driven QALNS/PPO) are systematically evaluated across six performance dimensions. Results show MILP establishes the theoretical optimality bound (1 582 CNY) in baseline scenarios but fails to solve within 1 200 s under sudden NFZ constraints. MSHOA yields the best solution quality (1 648 CNY) in conventional settings, though performance deteriorates under high-frequency emergency insertions. Notably, PPO demonstrates exceptional robustness across all dynamic scenarios, achieving a 0.4 s replanning time and > 95% task satisfaction after 10 consecutive emergency insertions, exhibiting the strongest robustness under tested conditions. Conversely, ACO and GA path feasibility plummets (23.8% and 44.9%) upon NFZ activation, exposing limitations in knowledge transfer under coupled constraints. This study pioneers a comprehensive benchmarking framework, providing a quantitative decision-making basis for robust low-altitude medical network scheduling.