Thesis/Capstone
Publication Date
Authored by
Abhinav Goyal
Advisor(s): Matthias Winkenbach
Topic(s) Covered:
  • Transportation
  • Urban Logistics
Abstract

The growing urban population over the past few years has created many challenges for last mile distribution, such as traffic congestion, pollution, and lack of parking space availability. With the advent of e-commerce industry, the volumes for last mile delivery are growing consistently. As firms struggle to provide competitive delivery commitments to the end customers, they are exploring alternative delivery methods, such as drones and e-cargo bikes to navigate urban areas efficiently while addressing pollution and traffic concerns. However, range and capacity constraints associated with such alternative delivery modes restrict the operations that can be carried out with such vehicles. Hence, firms are re-designing their last mile distribution strategies to adapt to the constraints posed by these delivery modes. One such strategy is to deploy a multi-echelon distribution network, using satellite nodes near customer locations that allow for transshipment. The large conventional trucks deliver parcels to a satellite, from whereon the parcels are cross-docked into lighter vehicles (such as e-cargo bikes) which perform the final delivery. This project introduces a mixed integer linear programming model for a two-echelon delivery network, to determine the optimal count and locations of satellites for a large parcel company. The transactional data for deliveries and pickups associated with one parcel center has been used to develop and test the model. The first-tier transportation, that is from parcel center to satellites, has been designed as a location routing problem. The second-tier transportation, that is from satellites to customer delivery points, has been designed as an allocation problem. The model tries to minimize the associated costs with the satellite operation, optimizing the fixed cost of establishing a satellite against the cost of distance traveled and transit time. Traffic considerations and road network distances have been accounted for by using real road network data for distance calculations, and transit times adjusted for traffic conditions across multiple hours during the day. The model returns the count of satellites to be established, along with their respective locations and vehicle routes for first tier transportation. Finally, the model maps all the customers to their respective satellites to achieve an optimum distribution cost.