Thesis/Capstone
Publication Date
Topic(s) Covered:
  • Optimization
  • Transportation
  • Last Mile
  • Urban Logistics
Abstract

With the latest technological advancement, the use of drones has emerged as an innovative and viable business solution for last-mile distribution. An efficient drone delivery system has to address the classic vehicle routing problem (VRP): "What is the optimal set of routes for a fleet of drones to serve a given set of customers?." The goal of this research project is to evaluate the optimal design and operational performance of four different drone delivery systems, using real-life last-mile truck delivery data. The authors quantitatively model four different drone delivery systems, from a pure drone delivery system to an unsynchronized drone-truck system and compare their relative benefits and shortcomings under various scenarios. A Memetic Algorithm, an extension of a Genetic Algorithm, is developed and used to optimize delivery routes of truck and drones for all the four delivery models.

Our research shows that Memetic Algorithm is quite robust handling VRP with 50 customers, yielding only 3.7% gap from the optimal solution. Among the four considered delivery models in this research, the Delivery System model 4 - where truck and drone share same area of service - performs superior than other three models, providing 100% coverage to all customers and reducing minimum tour time as high as 80%. The outcome of this research will help shape the quantitative and qualitative comparison of drone delivery systems and set the foundation for modelling and analysis of more advanced systems (e.g. synchronized truck-drone delivery system). It also helps industry to understand the possible use cases for drones in last-mile delivery and the most crucial levers of these models to maximize the performance of such drone delivery systems.