Robot-aided electric vehicle routing problem with lockers and prime customers prioritization

Abstract

Satisfactory and fast customer service is one of the critical parts of last-mile delivery. Companies like Amazon prioritize Prime members with same-day delivery while offering lockers for customer convenience. Additionally, robot-aided Electric Vehicle (EV) delivery is recognized for its cost efficiency and fast service in densely populated areas. Integrating EVs, delivery robots, and lockers, and prioritizing Prime customers can improve efficiency and service responsiveness. This integrated approach offers home delivery by EVs and robots and self-pickup from lockers. Every customer is assigned a prize (profit), with a higher profit associated with the Prime membership. Each EV dispatches robots, with a “dispatch-wait-collect” tactic, to serve the customers, while some customers are allocated to the lockers. This study introduces the Robot-Aided Electric Vehicle Routing Problem with Lockers and Prime Customer Prioritization (REVRP-LPCP), which aims to determine the least-cost routes for EVs and robots, assign customers to lockers, and prioritize prime customers by serving them within a single-period planning horizon. The REVRP-LPCP is formulated using a mixed-integer linear programming model, improving the EV-only-based delivery system by 52.94% and 21.95% in EV route and utilization costs on average. A metaheuristic is introduced, incorporating problem-specific repair and improvement operators to efficiently address large instances of the problem, outperforming Gurobi in 36 large instances by an average of 2.79% in terms of solution quality. Also, our method has identified 44 new best solutions in the related benchmarks. A comprehensive sensitivity analysis is conducted, assessing various scenarios and providing managerial insights. • New multi-modal delivery system via EVs, robots, and lockers for prime customers. • Problem is formulated by MILP model, compared with EV-based delivery. • A metaheuristic with problem-specific operators is proposed for large instances. • Our method outperforms Gurobi, obtaining 44 new best solutions in related problems. • Sensitivity analysis, model evaluation, and managerial insights are presented.