Energy-Aware and Delay-Sensitive Management of a Drone Delivery System

• 成果类型: Article
• 署名作者: Liu, Weiliang; Sun, Xu
• 署名单位: National University of Singapore; State University System of Florida; University of Florida
• 刊物名称: M&SOM-MANUFACTURING & SERVICE OPERATIONS MANAGEMENT
• ISSN/ISSBN: 1523-4614
• DOI: 10.1287/msom.2021.1056
• 发表日期: 2022
• 页码: 1294-1310
• 关键词: Dynamic Programming; queuing theory; service operations; stochastic methods; transportation; diffusion analysis; stochastic optimal control; drones

摘要

Problem definition: This paper considers a drone delivery system that delivers packages to multiple locations using a model that captures three distinctive features of interest: first, a battery-operated drone that can fly a limited number of distance units on a full battery; second, a set of demand locations with different flight distances and service-level requirements; and third, an adjustable flight speed to trade off energy efficiency for faster deliveries. By leveraging drone speed, job sequencing, and admission control, the system operator strives to achieve two managerial objectives: (i) to minimize energy- and congestion-related costs and (ii) tomaximize the use of the battery energy between consecutive battery replacements. Academic/practical relevance: E-commerce and shipping companies worldwide are testing and launching commercial drone delivery services to residences. According to a recent study, such services are expected to grow to $100 billion market by the end of the next decade. Our work focuses on improving drone delivery operations, which can reduce costs and improve user satisfaction. Methodology: Focusing on the first objective, we approximate system dynamics using diffusion processes and derive an optimal control policy for the approximating system. Results: Interpreting that policy in the context of the physical system while taking into account the second objective, we devise an implementable set of speed control, job sequencing, and admission strategies. The sequencing strategy, which selects a portfolio of jobs (termed as an activity) to be accomplished within a full battery cycle, results in a win-win scenario-it allows both objectives to be approximately optimized at the same time. Managerial implications: Using the activity-based prioritization scheme provides opportunities to improve the overall performance of the system. Additionally, to achieve the intended job prioritization, the system operator can rely on a small set of activities that are of equal size to the number of job classes.