Portuguese Logistics Association acknowledged route planning model developed by INESC TEC

The TacitRouting project – combining data science methods with traditional optimisation techniques for last-mile route planning to make them more efficient – received the PEL Academia 2025 prize by APLOG, the Portuguese Logistics Association. This award recognises academic initiatives with high impact and innovation that are applicable to the business and logistics sectors.

The last mile is the most critical and expensive sector of the supply chain. “Some studies suggest it can account for up to 53% of total transportation costs,” explained António Galrão Ramos. According to the INESC TEC researcher, this is the section where traditional optimisation methods show their clearest limitations. “Traditionally, systems calculate routes by optimising theoretical distances and times, but in practice, drivers often follow alternative routes. With our system, we’ve taken a decisive step towards finally mixing traditional optimisation with drivers’ implicit knowledge,” he stated.

The TacitRouting project, developed at INESC TEC, goes beyond traditional operations research methods by proposing a data-driven route planning model enhanced by machine learning approaches, to incorporate drivers’ expertise into the route planning process. This solution addresses one of the “most persistent challenges in routing: the systematic gap between theoretically optimal solutions and real operational efficiency,” said the researcher.

This Vehicle Routing Problem involves determining the optimal set of routes for a fleet of vehicles to serve a group of customers. However, when it comes to last-mile delivery optimisation, there’s a clear divide between the theoretical planning of routes and their real-world execution. For example, drivers tend to deviate from the originally planned route sequences to follow ones they find more convenient, since the quality of a route is not defined solely by theoretical duration or cost, but also by factors like geography, infrastructure, and customer specifics.

“Experienced drivers possess tacit knowledge about the complex operational environment in which they serve customers. They know which roads are difficult to navigate, where and when traffic occurs and parking is easy to find, which stops can be conveniently grouped, and many other factors that are difficult (if not impossible) to formalise in an optimisation model.”

According to António Galrão Ramos, leveraging learning-based approaches to understand and anticipate these deviations can lead to more efficient and safer routes, more sustainable delivery operations, greater driver satisfaction, and improved service quality. The solution developed by INESC TEC incorporates a multi-objective hierarchical districting system, which goes beyond traditional geographic districting criteria. “This system simultaneously considers demand patterns (e.g., seasonality), natural and artificial obstacles, and drivers’ knowledge,” the researcher claimed.

TacitRouting received the PEL Academia 2025 prize by APLOG, a “seal of quality” that will help open doors for the team behind the project. “This award validates that we are on the right path. Academic research can sometimes become detached from the real needs of the market. The fact that professionals in the field recognise the practical value of our work confirms that we’ve developed something truly useful,” the researcher emphasised.

In addition to António Galrão Ramos, the TacitRouting project team also includes INESC TEC researchers Farzam Salimi, Pedro Rocha (lecturer at ISEP), Manuel Pereira Lopes (lecturer at ISEP), Carlos Ferreira (lecturer at ISEP), Beatriz Santos, and João Silva.

The PEL Academia 2025 award aims to reward projects that develop and promote knowledge and innovation in the logistics sector, with business applicability, research relevance, and a high degree of innovation.