Thermal-Aware Particle Swarm Optimization for UAV Medical Delivery Route Planning in Hot-Climate Urban Environments

Abstract

Unmanned aerial vehicles (UAVs) can bypass the road congestion that delays urban medical logistics, but their use for temperature-sensitive cargo in hot-climate cities remains unproven: existing route planners optimize time, distance, or energy while ignoring the payload’s thermal state, and no published UAV energy model is calibrated above 40 °C. This paper formulates UAV medical routing in extreme heat as a four-objective minimization problem    delivery time, thermally corrected energy, a Payload Integrity Index (PII), and JARUS SORA v2.5 ground risk    over a verified 21-node Baghdad hospital network at 48 °C ambient, and proposes a Thermal-Aware Particle Swarm Optimization (T-PSO) algorithm to solve it. T-PSO is a memetic extension of discrete PSO combining greedy initialization with 2-opt local search, paired with a Thermal-Aware Intelligent Decision Engine (TA-IDE) that issues an ACCEPT, REPLAN, or REJECT decision under a hard payload-admissibility constraint. T-PSO is benchmarked against Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Ant Colony Optimization (ACO), Grey Wolf Optimizer (GWO), and Whale Optimization Algorithm (WOA) under an equal budget of 6,000 fitness evaluations across five scenarios, three weight profiles, and five mission types, giving 13,500 deterministically reproducible runs. T-PSO achieved the lowest mean composite fitness (0.1023), significantly outperforming GA (0.1052), ACO (0.1058), GWO (0.1111), WOA (0.1118), and PSO (0.1225) (Friedman χ²(5) = 143.66, p = 2.99 × 10⁻²⁹), with its largest margin    15.7–57.0%    on the 20-node full-network scenario. A component ablation attributes essentially all of this advantage to the 2-opt local search rather than to greedy seeding, and shows that a spatially uniform thermal correction cannot influence route construction at all; thermal awareness therefore enters the method through route evaluation    energy accounting, payload admissibility, and the TA-IDE    rather than through route construction. Sensitivity analyses over the thermal coefficient (5.7–17.1% energy overhead), container insulation class (τ = 31–1,527 min), ambient temperature (35–52 °C), instance size (5–20 nodes), and objective weights (T-PSO gain 9.5–19.9%) establish the boundaries within which these conclusions hold. The framework gives healthcare logistics operators an auditable, safety-aware routing basis for hot-climate UAV medical delivery. All results are simulation-based; physical flight validation is identified as the necessary next step and a protocol for it is specified.