Algorithm predicts problems in self-driving car tests

How self-driving vehicles should be allowed on public roads is a topic that is regularly discussed. But self-driving technology not only makes the use of cars on public roads different, but also tests of such vehicles. A new algorithm predicts when components in a self-driving car threaten to break down. This allows the vehicle to reach a safe and controlled stop.

The algorithm was developed by the Swedish Technical University of Chalmers. The university wants to contribute to the safe testing of vehicles, even if there is no human driver present in the vehicle.

Recognize signals in time

Testing cars is currently the job of test drivers. An important skill of these drivers is timely recognition of signals indicating that a significant portion of the vehicle is in danger of breaking down. This way, test drivers can anticipate this, bring their vehicle to a stop in time and prevent accidents.

However, self-driving vehicles are not always tested in the presence of a human driver. It is therefore important to notice such signals in a different way. Think of vibrations, deviating measured values, rising temperatures and other factors.

Test in the same place as ordinary vehicles

This is especially important as it is expected that in the future self-driving vehicles will be tested on the same test tracks as ordinary vehicles with a driver. For the safety of these drivers, it is important that problems with self-driving vehicles are detected in good time.

The research project is called ‘Enablers for testing autonomous vehicles at existing testgrounds’. The research maps the challenges that arise around testing vehicles on existing test tracks as cars become more and more autonomous. Researchers from Chalmers University of Technology collaborated with the truck manufacturer Volvo and the car manufacturer Volvo Cars.

Real-time warning system

The parties are now presenting an automated system that warns in real time if a mechanical component that is important for the safety of the vehicle is in danger of failing. “The problem lies in how to best monitor a vehicle if you do not know in advance what is going to break,” explains Tomas McKelvey, head teacher of signal processing at the Faculty of Electrical Engineering at Chalmers University of Technology.

Self-driving vehicles on the road are equipped with all kinds of safety systems. McKelvey stresses that during testing of vehicles, one cannot rely on systems whose presence on public roads is mandatory. This is because vehicle testing often takes place in the early stages of vehicle development. Such autonomous functions are then not always reliable.

Changes in vibrations

People discover problems based on, among other things, changes in vibrations. The algorithm works in the same way. Using accelerometers at various points in the vehicle, the algorithm detects the vibrations in the engine and the way in which these vibrations spread through components and the body. “Our hypothesis is that the nature of the vibrations changes if a problem occurs on board, and that these changes can be detected using instruments,” McKelvey says.

Of course, vibrations can also change due to other factors. For example, due to variations in road surface. A series of sensors located near the wheels therefore measure the vibrations entering the vehicle. The other instruments look for changes in vibrations in the vehicle.

Collect data with test vehicles

Collection of the right data played an important role in the development of the algorithm. Two test vehicles were used for this: a truck and a passenger car. These were tested on a Volvo test track in Hällered. This is also where the AstaZero test facility at Chalmers and RISE Technical University is located.

Based on data collected here, a mathematical description has been made of how the different parts of the vehicle move together. Based on this data, a model has been created that describes the normal situation. Each value may have some deviation.

Deliberately creating technical problems

The researchers then drove the two test vehicles over the same surface area as before. To test the algorithm, they deliberately created technical problems in the vehicle. The system proved to be quite easy to detect these problems. Even when it comes to problems, as test drivers say, are very difficult to detect.

In one case, the algorithm could not detect a problem. It was a nut that had come loose a little on the truck’s shock absorber. “But this was a problem that the test drivers could not notice either,” McKelvey said.

The nature of the problem was not found

At this point, the system only detects that a problem has occurred. The algorithm can not figure out what this problem is. The researchers expect this to develop, but did not have enough time for this in this study. McKelvey: “From a safety perspective, the most important thing is to detect that a problem has arisen so that the vehicle can be taken off the test tracks.”

McKelvey reports that there is great interest in the project from other test facilities. He hopes that the method will be further developed to an international standard for test circuits.

More information about the algorithm is available here.

Author: Wouter Hoeffnagel

Leave a Comment