DATAREL: project update

The Netherlands is a leading country in the field of transport, logistics and supply chain. However, the logistics sector is currently undergoing profound changes. The impact of e-business, big data and new technology is enormous. The goal of DATAREL (big DATA for REsilient Logistics) is therefore to develop innovative methods and techniques for data collection and analysis in the logistics sector, and to show how they can contribute to controlling increasingly dynamic and complex logistics processes. We spoke with Martijn Koot and Rob Bemthuis (both University of Twente) to get an update on this project.

Smart collection...

The DATAREL project started in 2018 and includes several work packages, covering the whole process from data collection to concrete changes in logistics processes. Rob works mainly on the data collection side, where the question of the relevance of data plays an important role. "A sensor can transmit the temperature of a pallet every second," he explains, "but is that useful? When is a disruption big enough to pass on to a higher decision level?" The term disruption is a central concept in DATAREL: the researchers use it to indicate a relevant fluctuation or unexpected and undesirable change in a variable - for example, a temperature fluctuation that can compromise the quality of a product.

Project partner Ahrma produces smart pallets with all kinds of sensors for determining, for example, shock, location and temperature. But it's not just sensors: the pallet contains intelligence and can reason for itself. In other words, the researchers are working on smart pallets that have a decision model on board themselves. Depending on the product on the pallet, the model applies certain rules, allowing the pallet itself to determine when a disruption is relevant and should be forwarded. For example, a pallet can detect that a load of ice cream has become too warm. In such a case, action is taken: the pallet can be written off immediately, which saves further costs and unnecessary movements in the transport chain.

Rob: "Actions that are necessary after a disruption are now all determined and deployed by human planners. This causes delays, and it is also difficult for a planner to make the right decisions based on limited and possibly unreliable data. Many of the choices planners have to make can be approached well with a logic model or with math. The best way is to use the computer for the heavy computational work and humans for more creative tasks. What's interesting is how you bring the two back together: for example, a model can give a suggestion of the 5 best solutions, then the human makes the trade-off and decides which one to implement."

... and smart use

The development of data-driven algorithms is Martijn’s area of expertise. "With sensors it is relatively easy nowadays to know all sorts of things about a product at any time of the day. How can we now use that data to make better choices, set up more efficient logistics processes and solve disruptions more easily, or better yet predict and then prevent them?" He goes on: "In an ideal world, every transport goes as planned and you don't need sensors. However, the world is not ideal: problems can arise at very many places in a logistics chain. Logistics planners have a wealth of experience and are constantly 'putting out fires'. That experience is of great value: our research therefore focuses not on complete automation, but on smart software that puts out a lot of small fires so that planners have their hands free to tackle the problems where human input is important." Rob adds: "The decision models we develop must additionally be able to deal with unreliable and incomplete data."

The running aground of the Ever Given ship in the Suez Canal was a sensitive tick for global trade but very interesting for the researchers, Martijn explains: "In the port of Rotterdam, they built a digital platform from scratch in one month to be able to manage the arrival of the "traffic jam" of ships that was stuck behind the Ever Given. The fact that this could be done so quickly indicates that the basic infrastructure is already in place. This applies to a very large part of the logistics sector: there is a world of data out there, but its use is still far too limited. With smart software, you can detect signals before things really go wrong and quickly generate options to solve problems."

Testing in practice

Rob recognizes that already available data is still used to a limited extent. "Many companies struggle with the question of what they can do with it. It's up to us to indicate what works and what doesn't: both are fruitful outcomes of our research. We work in DATAREL with concrete use cases from project partners, each with their own focus, such as Locus Positioning, Ahrma, Innovadis, Cape Group, Ovis Telematics, and Datacadabra. A number of partners already knew each other: this bond of trust certainly benefited the project. At the same time, due to corona, companies sometimes had other priorities - perhaps we weren't always able to get the most out of it."

Testing in practice is exciting, Martijn explains. "Innovations really do have to prove themselves in the competitive world of logistics. And we certainly don't want the results to deteriorate unintentionally. That's why we first do simulations, but I always look for a connection to real-life situations. Feedback from the project partners is important in this regard."

Drag-and-drop

From DATAREL, the researchers have contributed to a number of other interesting projects, including the OpenTrip model. This initiative is supported nationally and may eventually become the legal standard for data collection in the logistics sector. However, the main line of DATAREL is formed by the use cases from practice. At the end of the project, we would like to have a toolbox for practical and academic issues that can be used to make logistics processes more resilient. We are strong advocates of open science, so we will make as many methods, datasets and models (including source code) publicly available as possible." Martijn adds: "Ideally, we develop our algorithms to the point that it becomes 'drag-and-drop', so very accessible to use. The user gets a shopping list for required data, puts it into the algorithm and then gets recommendations to improve processes. Before we get to that point however, there is still some work to be done..."