Hybrid drone/truck last mile delivery planning for e-groceries
This PhD project focuses on optimising the routing and scheduling decisions for last-mile grocery delivery. The pandemic combined with the need to reduce greenhouse gas emissions by switching to electric vehicles make drones an attractive option for last mile delivery. The work will look into how best to combine trucks and drones in a home delivery operation for groceries.
It will also consider different home delivery platforms to assist decision-makers to opt for the optimum and near-optimum routes and schedules. A mixed-integer mathematical model will be programmed to obtain optimal solutions for small network instances.
Considering the large network of customer locations and local supermarkets, this project will also develop novel algorithms to solve the optimisation problem in a reasonable amount of time.
Participants
Project background
One challenge in using drones for grocery deliveries is posed by the weather. Consequently, the mathematical model under development will consider wind direction and speed.
A second challenge is the urgency of grocery delivery. A requirement for a quicker delivery will incur a higher cost. The mathematical model developed in this project will allow the study of the relationship between delivery urgency and system cost.
The third challenge is the weight limit that applies to commercially available drones, which may require multiple drone deliveries to the same address or alternatively delivery by truck.
Project objectives
In this study multiple trucks, each of which is equipped with drones, will via simulation deliver grocery purchases in parallel to customers. Routes, schedules, and depot locations will be sought, allowing the drones and trucks to fulfill the delivery tasks with the least cost, taking delivery time windows into account.
In order to maximise the satisfaction of customers, this study will develop a mathematical model that allows the minimisation of the time at which the last delivery task is performed (the ‘makespan’), considering the time window and other constraints.
Optimal solutions for the mathematical model will be found by standard solver for smaller problem instances and customised heuristics for larger problem instances. The sensitivity of system cost to different model parameters will be studied.
December 2022: Project completed
This project has been completed.