A Multi-Objective Optimization Model for Collaborative UAV–Rider Meal Delivery Routing Using an Enhanced NSGA-II Algorithm
Abstract
Urban instant-delivery platforms increasingly require efficient, punctual, and low-carbon delivery services. Unmanned aerial vehicles (UAVs) can reduce dependence on congested ground traffic in meal delivery and further improve overall delivery efficiency through coordinated operations with ground riders at rendezvous points. However, existing studies mainly focus on ground-based routing or simplify UAV-assisted delivery as a single-objective problem, limiting their ability to balance cost, completion time, carbon emissions, and service quality. To address this limitation, this paper investigates a collaborative UAV–rider meal delivery routing problem and formulates a multi-objective optimization model integrating restaurant pickup, UAV transfer, rider last-mile delivery, and soft customer time windows. An Enhanced NSGA-II algorithm is then developed, where hybrid initialization improves solution quality and diversity, adaptive operators balance exploration and exploitation, local search refines route structures, structural repair maintains feasibility, and diversity preservation supports a well-distributed Pareto front. Comparative experiments against NSGA-II, MOPSO, NSGA-III, MOEA/D, MODE, and IMODE, together with scalability, ablation, sensitivity, and case analyses, show that E-NSGA-II provides stronger Pareto-front approximation. The results support its use as a decision-support method for service-aware and low-carbon UAV–rider meal delivery, while also revealing additional computational cost.