Last-mile network design for urban commodity distribution in Latin America

Abstract

Transportation, as the carrier of freight and passengers, is undeniably one of the fundamental components required for both economic growth and development. In an urban environment, freight movements support most city-based activities, while detrimentally impacting the quality of life through negative externalities (e.g., congestion, noise and air pollution, etc.). Specifically, last-mile delivery is regarded as an important yet highly expensive section within every supply chain. This is partially caused by inherent inefficiencies such as prolonged delays in traffic and unproductive idle periods at customers locations, among others. Consequently, there is a need for methodologies addressing the design of improved last-mile delivery networks. In this context, the optimal design of distribution systems requires an integrated view of strategic, tactical, and operational decisions. This work contributes with a mathematical framework that provides such an integrated view while leveraging both customer-generated waiting time inefficiencies and existing network infrastructure to serve additional clients. It also provides computationally feasible algorithms to obtain solutions for realistic situations. First, we formulate a single-echelon, multi-depot, capacitated routing problem. Employing a brownfield approach, this model optimizes the fleet composition as well as the delivery schedule and allocation to distribution facilities of medium- and high-dropsize clients, hereafter 'big-box' customers. This Routing Problem (RP) is modeled as a special case of a Bin Packing Problem (BPP) combined with a customer clustering approach. However, given its high combinatorial complexity, two alternative methodologies, a two-step approach and Benders Decomposition (BD), are tested to reduce computational times. Second, we develop a two-echelon extension, which builds on the previous model, to evaluate the economic impact of including a large number of low-dropsize customers, also known as 'nanostores', into the original distribution footprint. Those newly added customers will be served through the second echelon using a subset of the original big-box customer locations as transshipment points. To solve this Location-Routing Problem (LRP), a three-step iterative optimization approach is developed and tested. Both models are applied to a real-world consumer goods distribution case study in Latin America. Results suggest that a systematic and properly framed optimization approach, which makes efficient use of available resources, can significantly reduce the total distribution cost. Further, we show that the case study company, leveraging its existing assets and addressing inherent network inefficiencies, can efficiently expand its distribution footprint towards nanostores.