United States-European Commission Urban Freight Twinning Initiative: Compendium of Project Summaries
Overview of Second Annual Urban Freight Roundtable at 2017 Transportation Research Board Annual Meeting

---

Research

United States

Distribution Network Design for E-commerce in High-density Urban Areas

The purpose of the project was to create a large-scale, numerical optimization model for the design of urban distribution networks for major consumer markets of our collaboration partner, B2W Digital of Brazil. The model addresses:

• The optimal number, capacity, type, and location of B2W Digital's urban distribution facilities.
• The optimal allocation of service areas to these facilities.
• The location-specific optimal choice of the vehicle type and delivery model to be used to serve the customer.

To depict urban complexities as accurately as possible, the model was designed to incorporate various sources of relevant (big) data such as: company order and delivery data; high-resolution GPS traces from the current delivery fleet (in photo below); publicly available data such as travel times and distances from Google Maps; and geo-referenced information on real-estate cost and crime incidents.

High-resolution GPS traces to inform model parameters for travel speeds, traffic patterns, stop locations, and durations.
Source: MIT Megacities Logistics Lab.

Project Type

Research

Period of Performance

February 2016 - January 2017

Project Site(s)

São Paulo, Brazil

Website

megacitylab.mit.edu

Contact

Matthias Winkenbach
Director
MIT Megacity Logistics Lab
(857) 253-1639
MWinkenb@mit.edu

Challenges Addressed

• Efficient urban distribution of e-commerce deliveries.
• Design and planning of distribution network.
• Use of public and commercial data to improve last-mile efficiency.

Key Accomplishments

MIT produced a data-driven network design tool to be used by B2W Digital in its strategic design and operational planning of urban last-mile distribution to consumers in major Latin American cities. The tool allows B2W Digital to operate:

• At lower operational cost.
• With reduced impact on the environment.
• At a higher service level to the customer.
• With reduced risk exposure (i.e., likelihood of robbery attempts) for employees.

MIT developed novel methods of incorporating large amounts of high-resolution data from various sources into improved, large-scale, mathematical optimization models for the design and planning of distribution networks. In particular, MIT used the following data sources to better inform models of numerical network optimization and geometric probability methods for route length and cost estimation:

• Corporate transactional data.
• Fleet GPS data.
• Google Maps application program interface (API) data.
• Geo-referenced datasets of safety incidents and real estate cost.

Stakeholder Involvement

Private-sector research partner: B2W Digital, a major Brazilian e-commerce company.