Algorithms pave the way for sustainable road transport

The founders of Opt1dev, originally focused on route optimization for passenger and target group transportation, discovered that their advanced algorithms can also be effectively applied to freight transportation. This leads to significant efficiencies, with savings of up to 20 to 30 percent in kilometers and service hours compared to manually planned routes.

Your model is to make road transport more sustainable. Can you tell us something about this?

Frank: "With the algorithm in our planning engine, transport companies can save on fuel and the number of vehicles they use. They also save on service hours for drivers and planners and have a schedule that statistically best matches the order commitments. Because it also saves on mileage, and thus fuel, it also contributes to making transport more sustainable.”

Arthur: “Our passion is optimizing transportation, and in many cases that automatically leads to sustainability in the form of lower CO2 footprints. Personally, it really does make me happy when we can contribute to sustainability.”

What aspects, among others, does the model rely on?

Arthur: “The most important for us are the kilometers saved. That's also where you see the sustainability aspect most clearly.”

Frank: “Another aspect is vehicles. What vehicles should you deploy to operate as optimally as possible? This also means we are increasingly looking at what percentage of trips we can schedule on zero-emission vehicles.”

Arthur: “And then there are load factors and the specific requirements of orders. If you can apply more flexibility within an order, you can combine more efficiently.”

And when it comes to dynamic aspects, traffic situations for example?

Arthur: “An accident, for example, is of course unpredictable. Road works are. That's something we can take into account. The same goes for recurring traffic jams. Everyone knows that if you have to be in the center of Rotterdam at 9 a.m. in the morning, you must reckon with a longer travel time.”

How does the model work in practice?

Arthur: "We offer this as 'planning as a service'. We have a basic algorithm that solves a Vehicle Routing Problem with Time Windows (VRPTW): a well-known challenge around finding the most efficient routes for a number of vehicles, given a set of customers to visit at a given time. In our model, this includes commodity characteristics, loading/unloading aspects and fleet. For each use case, we build a kind of custom subsolvers here. Below this is a generic attribute layer, which allows you to seal off all combinations that are and are not allowed. These are the so-called constraints. Depending on these constraints, we build the solvers. Concrete examples are the time windows, or where the truck must end at night, because the truck must be reloaded that same night."

How much efficiency gain can the model bring transportation companies?

Frank: "Simulations show an efficiency increase of 20 to 30 percent in mileage. In practice, of course, you may encounter constraints that require adjustments, but if you're left with 10 percent efficiency, that's still huge. Besides, transportation optimization is becoming an increasingly difficult task for humans to perform. This is due to such things as environmental zones, time windows, specific attributes and other requirements. So manual planning is becoming increasingly complex, which also makes it more difficult to do it in an efficient way."

How do you see the role of planner in the future?

Arthur: "When we work with planners, we assume that the planner still has his own planning. He can compare that planning with the automated planning in our application. Eventually, the intention is for the planner to switch to automated planning. However, the planner always remains "in charge. For example, he can always override the automated schedule. When this happens, we record that action, because we want to know how often it is overridden. If that action is something we need to take into account in the future, then we include that information in the optimization round."

Frank: "By the way, we don't often see that manual intervention makes the route more efficient. What we do see is that you can change the rules of the game through your own interpretation. If a planner knows that customer A's time window has more stretch than formally established, he or she can respond to that and positively influence the trip. That is the judgement of a human being, which is why I also call our automatic planning augmented planning. It's about the interaction between man and machine that leads to the very best result in reality.

We have also seen that it is very important to involve transportation company planners in the implementation of automatic planning. They are needed to determine what all that automatic planner needs to take into account. Moreover, planners may feel that they are no longer needed. Nothing could be further from the truth: we see the system as a tool for the planner. A tool that ensures that the planner no longer has to do the repetitive work ̶ the routine calculation and puzzling is taken away from him or her. This makes the role of the planner primarily more tactical.”

To what extent is the model transparent? Is the result always easy to explain?

Arthur: "With AI, certain outcomes sometimes remain 'underwater.' That plays much less of a role with us. In principle, we always ensure that the outcome remains explainable. However, because of the complexity we can never fully guarantee that our solution is optimal. In order to be 100% sure, the software would have to run for years. What we can do is provide the most optimal solution for the time span given to us. And then we certainly come close to a 100% optimal solution.”

Frank: "You can run the system for five minutes, and then you end up with, say, 96 percent optimal. You can also run it for fifteen minutes, so you end up with, say, 98 percent.”

Do you have any ideas on how to plan even more sustainably in the future?

Arthur: “If you think of transportation as a single system, you could drive even more sustainably, for example by taking a package from another company when you are in the neighborhood. However, this kind of cooperation is difficult to implement, because there is a lot of competition in transport.”

What do you ultimately want to achieve as a company?

Frank: “We want to become a European player specializing in optimization planning in freight and passenger transport. We see that as two main segments. And then we have the specials, such as transport optimization for gritters. That's a whole other line of business, because they have to make as many meters as possible.”

Arthur: “Specifically for freight, we want to start focusing on multi-day algorithms, i.e. automating across multiple days. An important and quite complex activity. We would also like to plan dynamically, as we already do in passenger transport. What is also important is that we are going to take the capacity of drivers and vehicles into account. Now we still consider these factors as a given, a constraint, but you can also plan the services and your fleet based on the orders and volumes you have.”

Frank: “The last thing I would mention is making an auction system. Such a system can be very interesting for a carrier. We have already developed such a secure system for Connexxion: the cab auction. It would be really cool to make something like that for freight transport as well.”

Centric and Opt1dev

Centric is enriching its Transport Management System (Plan&Go TMS) with Opt1dev's intelligent planning system. This allows Plan&Go TMS users to take a big step forward toward further optimizing road transport and the underlying business operations.