Study On Operational Policies And Configuration Optimization In Drone Delivery Systems

With the rapid development of e-commerce,online shopping has gradually become the most popular consumption pattern.The increased transport of goods in cities has brought great challenges to the logistics services industry.B2C e-commerce enterprises need to fulfill customer orders and complete home delivery services in a timely manner,and the timeliness of order delivery plays an important role for enterprises to maintain brand image and improve competitive advantage in the market.After Amazon proposed using drones for door-to-door delivery service in 2013,major e-commerce companies have also been experimenting and testing drone delivery systems.With rapid delivery speed and excellent flexibility of throughput capacity,drones are considered as the most promising means of delivering goods in the future.Establishing a drone delivery system requires high investment on drones and complex control system to operate a fleet of drones with safety and stability.A good understanding on the system performance is essential before the implementation of a drone delivery system.Drone delivery system is a relatively new research topic of facility logistics.Related studies mainly focus on the routing problem.It is hard to find research works that explore the operating policies analysis and configuration optimization on the systematic level of a drone delivery system.Operating policies and system configuration may have great effect on system performance and operating cost.Therefore,Researches on these problems will have significant theoretical reference value for the application and popularization of the drone delivery system.Based on these above,this dissertation studies the operational strategy and configuration optimization of drone delivery systems.First,this dissertation studies a multi-warehouse drone delivery system,considering the allocation rule that all warehouses share the drones and the allocation rule that each ware-house owns its drones.Both plug-in charge and battery swap strategies are investigated for battery management.This dissertation examines the random and closest drone to warehouse assignment rules,and design a heuristic to improve the drone to warehouse assignment rule.A closed queueing network is built to estimate the maximum throughput capacity and a cost minimization model is developed for cost analysis.This dissertation validate the analytical model by simulation and conduct numerical experiments to analyse the operating polices.The results show that the closest drone to warehouse assignment rule outperforms the random drone to warehouse assignment rule when the number of drones is not large,and our heuristic can improve the throughput capacity by about 13.31%.The battery swap strategy provides a better throughput capacity than the plug-in charge strategy in most cases,while it needs more investment.Moreover,the shared allocation rule gives a larger throughput capacity than the dedicated allocation rule,and it reduces the operating cost by about 30.70%.Then,this dissertation proposes a truck-drone delivery system that uses a fleet of electric trucks,in tandem with drones,to fulfill deliveries.In such a system,a truck departs from a warehouse,carrying a batch of parcels and drones,then traverses several stop points to deliver parcels by drones.The influences of the capacity of the drones fleet in the truck,the stop points on the path of truck and the demands allocation among stop points of the truck on the system performance are explored to minimize the system cost.This dissertation builds a closed queueing network to estimate the throughput capacity of the system and builds a cost minimization model for the system with a required throughput time.This dissertation verifies that the theoretical model can estimate system performance accurately and effectively by simulation.In addition,this dissertation carries out numerical experiments,and finds that the more drones loaded on truck,the greater the throughput of the system,but there is a ceiling.There is an optimal value of the number of stop points on the truck path when the number of drones loaded on a truck is small.And when drones are sufficient,system manager should choose the minimum number of stop points on path of truck under the premise of the drone flight distance.Truck loads parcels with random policy can provide better system then demand-dependent policy.When the number of drones loaded in the truck is certain,the system throughput is in direct ratio to the number of trucks.Third,this dissertation optimizes location of the wireless charging station for a drone delivery system.This dissertation proposes to use the wireless charging device of drone to install it on the top of commercial buildings as a wireless charging station for the delivery system,thus expanding the service range of the system.While previous researches do not use customer behavior data,this dissertation proposes new models,integrating with customer behavior data analysis,to optimize wireless charging station for e-commerce companies.A real customer behavior data that contains 522,256 records is explored,and useful futures are extracted from raw data.This dissertation designs an approach to balance the predictive and interpretable performance of a decision tree with model distillation and heterogeneous classifier fusion.The decision tree is then used to predict valid demand information,including potential customers and their purchase probability.Based on the obtained demand information,a multi-period location model is constructed to optimize the the wireless charging station with the goal of minimizing operating costs.In addition,a comparative experiment is designed to verify the methods designed in this dissertation,and the results are demonstrated by numerical experiments.Finally,the relationship between customer service level and operating costs is quantitatively analyzed,which provides theoretical reference for e-commerce enterprises to weigh them in location decision-making.At last,this dissertation explores a drone food delivery system which serves a city and contains several fast food restaurants.Three devices(such as drone,battery,etc.)allocation rules are proposed:equal rule,demand-dependent rule and service-range-dependent rule.This dissertation evaluate the system via semi-open queueing networks and apply throughput time and waiting time as measures.Simulation model validates the analytical model.Results of numerical experiments show that the service-range-dependent rule is the best when the number of drones is small,otherwise,the demand-dependent rule is the best.