Daniel Olejarz, MASc thesis, Civil Engineering, 2020
Professor Matthew J. Roorda, Supervisor
Summary
Last mile parcel delivery is a particularly costly element of the freight supply chain. This challenge is exacerbated for courier companies by the rise of e-commerce and the business-to-consumer (B2C) model. B2C e-commerce shipments are growing faster than traditional business-to-business (B2B) shipments but are more costly for couriers. Autonomous ground vehicles (AGVs) are an emerging technology that has the potential to reduce the cost of last-mile delivery.
AGVs are small autonomous robots that can carry multiple parcels in secured compartments on streets, sidewalks, or in large indoor spaces. This thesis quantifies the cost of a last mile parcel delivery system supplemented with AGVs and compares it to the cost of the current manual system. Descriptions of the autonomous delivery vehicles receiving the most investment in industry are first provided. Next, a literature review which focuses on automated parcel delivery and split delivery vehicle routing problems (SDVRPs) is presented. The literature review identifies recent studies on automated delivery systems which show that automation can lead to significant cost savings and environmental benefits. The literature review also highlights the applicability of SDVRPs to routing AGVs but recognizes the inconveniences for the customer and the courier of splitting deliveries between vehicles. A novel synchronized split delivery vehicle routing problem (SSDVRP) is formulated. This model introduces additional constraints to SDVRP to ensure deliveries split between multiple vehicles arrive at their destinations concurrently.
Four days of sample parcel data and details on cost for staff and AGVs are provided for analysis by a major courier company operating in Canada. Parcel demand data are described with summary statistics and visuals. A heuristic solution to SSDVRP is then applied to the sample data. Scenarios with different vehicle capacities are considered when analyzing cost implications of using AGVs. These results are compared with a baseline value for manual deliveries. The results show cost savings scale with the rate of customers that can be served by the AGV. Cost savings are observed for all capacity scenarios. Higher capacity vehicles increase cost savings, though the savings diminishes as capacity increases. These results are shown in Figure 1.
The results are contextualized for courier companies looking to implement automated delivery systems. Recommendations are made on how to best use human couriers and AGVs simultaneously in a way that creates synergy between the two delivery methods.
Supervisor contact information
Professor Matthew J. Roorda
Tel: (416) 978-5976
matt.roorda@utoronto.ca