AI-Driven Autonomous Drone Routing for Tuberculosis Medication Delivery in Remote and Underserved Areas

Reliable tuberculosis (TB) medication delivery in remote and underserved areas requires more than physical arrival; the transported medicine must also remain therapeutically usable under environmental and operational uncertainty. Existing UAV routing methods usually focus on path efficiency, obstacle avoidance, or delivery completion, while treating medical payloads as inert cargo. To address this limitation, this study presents AI-DRTB, a Dueling Double Deep Q-Network with Prioritized Experience Replay (D3QN-PER) for treatment-sensitive autonomous drone routing. The framework jointly models thermal payload degradation, cold-weather battery stress, wind, rainfall, communication quality, obstacle proximity, and payload-related energy burden. Performance is evaluated using the Clinically Effective Delivery Rate (CEDR), which distinguishes geometric arrival from medically useful delivery. Across 200 unseen episodes in a stochastic 1, km × 1, km environment, AI-DRTB achieves 98.5% delivery success, 75.5% CEDR, and a 1.5% abort rate, outperforming DQN (66.5% CEDR) and DDQN (5.0% CEDR). The results highlight the importance of treatment-preserving metrics for autonomous medical-drone logistics.

Hali tarjima qilinmagan