Defesa de doutorado do discente Vinícius Gandra Marins Santos, dia 05/07/2023.

Defesa de doutorado do discente Vinícius Gandra Marins Santos, dia 05/07/2023.

Título: Optimization algorithms for vehicle routing problems with multiple decision levels
Data: 5/7/2023
Horário: 17h00 (Horário de Bruxelas)
Local: Microsoft Teams link: Click here to join the meeting (Meeting ID: 394 376 552 570. Passcode: F9EKE5)

Resumo:
In their day-to-day operations, logistics companies typically need to address multiple decision levels in addition to a vast number of realistic problem features. These decisions are not only related to transportation, but can also involve determining the location of facilities, arranging items for shipment or scheduling personnel. Addressing all of these decision levels is crucial given that they have direct impact on resource usage, costs and customer satisfaction. Meanwhile, some examples of realistic problem features include time windows, multiple depot options and working regulations.

Developing solution approaches for these extremely complex problems is often time-consuming and can easily lead to a complexity explosion. In an attempt to help orient researchers and practitioners who face this challenging problem context, this thesis introduces a general methodology for addressing problems that feature multiple interconnected decision levels with an underlying vehicle routing problem. This methodology should provide those confronted with such problems a general set of steps they can take to simplify and accelerate the problem-solving process.

In order to arrive at such a general methodology, we investigated challenges faced by a diverse set of logistics companies and developed approaches to solve each problem. We investigated how the different decision levels interact with one another, how we can take advantage of such interactions, what methods can reduce and effectively explore the search space of the problem, and how we can minimize the risk of a complexity explosion in order to quickly produce high-quality solutions. Ultimately, rather than beginning by focusing on what sets these multi-level problems apart, we hope this thesis will encourage other researchers to first exploit what they share in common.