Smart car
​​​​
​genre picture from rawpixel.com

​This is how the next generation of self-driving cars will be equipped

​At the tech company Asymptotic AI, a research project is currently underway to create safer systems for self-driving cars and assistance systems. The project uses a car with six different cameras and a laser scanner that measures distances to objects. The car drives around Gothenburg and collects data for machine learning.
Yinan Yu is assistant professor at the Department of Computer Science and Engineering and one of the co-founders of Asymptotic AI, which has a close connection to the Chalmers research center.

”It is important to use several systems that complement each other for data collection and analysis, because the systems have different purposes and characteristics” says Yinan Yu.

”Perception, i.e. the brain's handling and interpretation of information, is the basis of the entire system. Because if you can't see, there's no chance you can make the right decision. For the perception part, you need two things: you need to see and you need to understand what you see”  she says.

Are the recorded videos saved afterwards, or are they only used while the car is being driven?

“The usual thing when you have self-driving vehicles that film the surroundings is that you save the recordings to use them for development purposes. The analysis of data is crucial for quality and very important for security. But larger volumes of data collection and further AI development are needed before self-driving vehicles can be used” says Yinan Yu.

“There are many scenarios that can become very critical in the automotive industry, where you have no room to allow yourself to fail. That's why self-driving vehicles haven't been put on the roads yet, we don't really have control over them” she says.

What are the risks of this type of project?

“GDPR is something that must be taken into account, for example. Filming and recording people on the street is clearly an invasion of privacy, but since we drive around with cameras on the car, filming people cannot be avoided. One solution could be to have clear information on the car that people are being recorded” says Yu.

“We have also constructed an anonymization software that blurs both people's faces and the number plates caught on film. But sometimes it's not enough to just blur people's faces, you can still recognize them based on their body shape, clothes and other things”

There are several anonymization programs that Asymptotic AI uses in its research and product development. The programs can for example erase an entire body, and according to Professor Yu, the development of new versions of these programs is a necessary step.

One-sided information and legislation create challenges

Something that is still not mapped out is what the consequences could be of collecting, analyzing and saving recordings of people out on the streets. The legislation surrounding this type of recording is not fully developed, leaving room for uncertainty.

Another challenge Yun sees is that AI easily builds bias into the data used.

“In addition to it being a question of having a representation that should correspond to reality, there are also security aspects linked to bias in the data. Problems can arise if the needs of all groups in society are not included when building a safe city and vehicle assistance system” she says.

What is the most interesting part of AI in the automotive industry?

“I think it's fantastic that technology develops so quickly. There are so many people nowadays working on the combination of AI and vehicles and it is really progressing” says  Yu.

Written by Agnes Ekstrand and Camilla Jara

Page manager Published: Mon 31 Oct 2022.