Respond to ever-changing situations using self-learning systems

Whether it concerns the efficiency of batteries in electric vehicles or the reliability of high-speed data communication, in both cases, unforeseen circumstances can interfere with performance. Therefore, better ways to respond to these ever-changing situations are being sought. Self-learning algorithms, which can combine existing knowledge and a large amount of collected data, play a crucial role in this work. Under Automatic adaptation research program, Dutch Organization for Applied Scientific Research (TNO) is exploring the potential of this promising innovative technology.

Frank Willems, a professor at Eindhoven University of Technology (TU / e) and a leading TNO scientist in the traffic and transportation department, along with Gert Witvoet, a university lecturer at TU / e and a senior TNO specialist in the industry department, spoke a year and a half ago. Willems: “I just created the AutoAdapt research program. It turned out that we have more in common than we thought. It’s a good thing to work at TNO – it’s that you deal with so many different disciplines. For example, this multidisciplinary research program involves colleagues from at least five different departments: Traffic and transportationthat I work in Industrial unitthat Gert depends on it as well ICT, Energy Transition and Defense. All of these people bring their own experiences. ”

Frank Willems © TNO

Self-learning systems

According to Willems, the goal of the research within AutoAdapt is to combine existing knowledge with data. “You can ask questions like,‘ How will the car do in ten years? How will it be used during this period? Despite the very different uses, is it possible to guarantee the lowest possible cost? ‘ To answer these and other questions, you need to do a lot of testing in the lab right now. ”

Adaptive systems, Willems said, make it possible to make good predictions not only about any conditions based on available data, but also about ways to predict them.

“By incorporating the results of other work we have done before, in addition to monitoring the progress of real-world demonstrations, these adaptive algorithms can also be studied. This allows us to significantly reduce the time required to develop and test and new ones can find their way to the market faster. The potential of these self-learning systems is huge. This is especially true for complex systems. However, we also see that the difference between laboratory conditions and industrial use is very large. eliminate. “

Subscribe to our newsletter!

Review your weekly news Every Sunday the best articles of the week in your inbox.

Demonstrate in practice

The aim is to show in real practice over the next few years how self-directed learning, adaptive methods and linking existing knowledge with qualitatively valuable data can be used to accelerate innovation.

One such promising innovation is technology optical satellite communications, through which an incredibly large amount of data can be transmitted through laser beams. Vitvoet: “Self-learning algorithms are useful here when configuring laser light. This guarantees optimal performance. In other words: high speed, but without compromising stability and connectivity.”

TNO
Gert Vitvoet © TNO

In the case of machines, especially electric and hybrid trucks, to which Willems ’research focuses, self-learning algorithms in turn ensure the right decisions are made. This can include charging times and battery charging profiles for both hybrid and electric trucks, but also on the route.

Willems: “The last thing you want is for the car to suddenly stop because the battery is empty. On the other hand, you don’t want the battery to run out too quickly. With the help of self-learning algorithms, cars can make good decisions in the future.” . That means you can put a lot of data into the cloud, as Tesla already does. “

Risk factors

The point that connects the Willesms and Witwoet research areas is that they both deal with unforeseen working conditions. Willems: “Think, for example, about changes in weather conditions or the flow of traffic. maximum work is required. For example, battery life depends on usage and charging profiles. When driving in hot or cold weather and the selected route has a strong impact on power requirements. Despite these differences, we want to to maximize the working time of a truck or bus.To transport as much cargo or as many people as possible.

TNO
© TNO

Even in optical satellite communications, Vitvoet adds, internal and external factors, such as atmospheric disturbances and vibrations from the artificial satellite used, as well as the individual characteristics and state of the individual and the receiver, can cause data distortion. “Vibration can affect the accuracy of the wood. If you have a narrower light, then this will allow faster data traffic, but at the same time it is more vulnerable to interference. On the other hand, if the stick is too wide, it’s because of the speed of the data. “

Ability to adapt

Now how can you make a programmer – a man who adjusts all the parameters to optimize everything – with self-learning and adaptive systems that are able to make changes based on existing knowledge and data to achieve the expected result?

”Willemes:“ You can equip machines with an operating system that is capable of making all sorts of decisions thanks to these self-learning and adaptive algorithms. This gives many advantages. As well as greater safety on the move, as well as optimal battery usage and in the case of hybrid cars, the engine is also fuel efficient. Along with all the associated positive effects, such as less nitrogen and CO2 waste. On the eve of 2030, we need to move to zero-waste domestic cities. Using sensible information about routes and traffic flow, battery status can make it even faster for hybrid cars. “

Vitvoet is also very optimistic about the possibilities of self-learning algorithms. “Equipping the transmitter and receiver with a self-adapting function in optical satellite communication allows them to adapt the configuration to changing conditions. And as a result, always get the most out of your optical connection. This ensures optimal and consistent data transfer. “

Primary research program

Williams explains that the AutoAdapt program is what is known as the Early Research Program, in which research is conducted across disciplines. Most importantly, this is research that usually precedes the actual implementation of the innovation. “But you don’t have to collect it as a group of researchers in a room behind a desk. It’s definitely research that’s in the early stages, but with a clear intention to use it in the industry.”

Point on the horizon

In the case of Williams, this is also the end point of the horizon: “We really want our understanding to produce results that can be demonstrated in practice through demonstrations. To this end, we would like to initiate a discussion with industry partners with the relevant user case. We are ready to present the results of our research in “real world” situations. For example, we already have a small demonstration unit on the motor campus in Helmond. Where, even though it is still in a partially virtual environment and with a virtual machine, we can show the battery performance in the real world. “

However, researchers love big dreams. Vitvoet: “The insights we are developing with AutoAdapt can also lead to huge savings in production costs and time. For example, by helping to create a situation where lower material requirements or manufacturing tolerances are possible. We look forward to working with our industry partners to reduce the cost of implementing our innovations. ”

The Internet is more stable

So what will we, as consumers, see in all of this? Williams: “Cars that can be produced faster and cheaper. With minimal operating costs for the user thanks to maximum energy efficiency, maximum usage and longer battery life. But the insights we are developing with self-learning algorithms can be used in other areas in the future. Such as the trucking industry, the energy industry, the semiconductor industry, or robotics. ”

Vitvoet: “Self-propelled systems can also provide better and more accurate laser communication with the satellite. And the Internet is more stable and faster, capable of transmitting thousands of times more data than it does today. Probably even 6G. What the automotive industry can also benefit in an instant! ”

Willems: “Well, this is just one example of the advantages of being part of such a multidisciplinary team. I can use the knowledge that Gert has at home. “

Vitvoet: “And vice versa!”

Leave a Comment