Full truckload pickup and delivery problem: optimization methods and practical constraints of the supply chain middle mile

• Grant number: 24/08848-3
• Support Opportunities: Regular Research Grants
• Start date: July 01, 2025
• End date: June 30, 2028
• Field of knowledge: Engineering - Production Engineering - Operational Research
• Principal Investigator: Antônio Augusto Chaves
• Grantee: Antônio Augusto Chaves
• Host Institution: Instituto de Ciência e Tecnologia (ICT). Universidade Federal de São Paulo (UNIFESP). Campus São José dos Campos. São José dos Campos , SP, Brazil
• Associated researchers: Cleder Marcos Schenekemberg ; Mauricio Guilherme de Carvalho Resende

Abstract

This project aims to study a rich variant of the full truckload pickup and delivery problem. The variant is based on the operations of the middle mile in the supply chain. It addresses a configuration where a logistics service provider needs to transport full loads between multiple distribution centers. The logistics operator has a limited fleet of vehicles and has the option of outsourcing part of the demand. In line with the growing environmental and ecological concern, a fleet of vehicles powered by electricity or fuels from renewable sources is considered. This implies the need to establish a plan for refueling vehicles at strategic points in the transport network. In addition to classic routing constraints such as time windows, pairing, and pick-up and drop-off points, the variant also handles scheduling the work of a set of drivers over a planning horizon. Each driver has bio-rhythm constraints, which indicate the start and end times of activities, as well as travel limitations, which prevent him from leaving his home for a maximum number of days. These constraints are particularly relevant in the context of middle-mile operations and must be considered when planning the fleet's itineraries. To solve the problem, a mathematical model and an API (Application Programming Interface) that integrates several metaheuristics in a collaborative and parallel framework will be developed. Such an API will be generic and can be reused to solve other combinatorial optimization problems simply by implementing decoding functions. As a methodological innovation, we intend to explore the concept of random keys, proposed in Random-Key Genetic Algorithm and in Biased RKGA, in several classic literature metaheuristics. The implemented algorithms will be evaluated in public datasets from the literature and in datasets generated specifically for the variant under study. At the end of the project, the API will be made publicly available to the academic community. (AU)