Safer self-driving cars require efficient algorithms. INESC TEC part of the path of future mobility

INESC TEC participated in the project THEIA – “Automated Perception Driving”, a partnership between Bosch and the University of Porto, which presented results in September. The project focused on the development of sensors to improve road safety, and the Institute developed neural network architectures to improve autonomous vehicles’ perception of their surroundings.

What if, in the future, we get into a car and the main concern is not the path we have to take to get home safely? The idea of a safer car, which turns the driver into a passenger, free to talk or read, while the car operates autonomously has been around for decades. But for that day to come, it is necessary to provide these vehicles with a better decision-making ability.

This was the focus of the THEIA project – which, for three years, has developed technologies to improve the sensory capabilities of autonomous vehicles, using perception algorithms that take advantage of sensory data collected. INESC TEC contributed to the aspect of “perception” – a key concept in the mobility of the future.

The main role of the set of sensors is to become a substitute for the human driver’s vision. How? By perceiving the environment surrounding the vehicle and providing information that allows immediate decisions while operating.

“INESC TEC’s contributions to this project focused mainly on perception through LiDAR technology (remote sensing technique that consists of measuring the properties of reflected light in distant objects) and video cameras, the development of efficient neural network architectures for hardware inference, explainable perception and transparency of deep learning algorithms, and safe computing in Cloud and Edge environments – considering self-driving vehicles” explained Luís Pessoa, researcher at INESC TEC.

One of the main challenges of perception is predicting behaviours and actions in dynamic scenarios. The sensors developed within the scope of this project can identify and recognise cyclists, pedestrians, and vehicles – and the results can contribute to road safety, both inside and outside the vehicle.

“The goal was also to improve the decision-making ability of autonomous vehicles through Artificial Intelligence, while integrating cybersecurity to protect system integrity and user privacy”, added the INESC TEC researcher.

The results of the THEIA project were showcased during a public demonstration session that took place on September 25, at the Alfândega do Porto. The project “successfully achieved the expected objectives” and represented an overall investment of €28M; it brought together 250 professionals, to meet the goal of 33 technical scientific papers and the submission of 10 patent applications.

The THEIA project was co-financed by the European Regional Development Fund (ERDF), through the Operational Programme “Competitiveness and Internationalization” (COMPETE 2020) and Portugal 2020.