The project is carried out by a consortium led by the Motor Transport Institute and the Faculty of Electronics and Information Technology of the Warsaw University of Technology.

Aleksandra Rodak, International Cooperation Coordinator at the Autonomous and Connected Vehicles Competence Centre, said the database creation is a multi-stage, complex process, conducted partly on the roads and partly in computer laboratories.

“First, we need to identify particularly challenging road areas. We will visit them in a car equipped with a battery of various sensors. In traffic, we record short sequences of passages through specific sections of infrastructure – we record signs, traffic signals, and problematic situations, such as police directing traffic or the passage of an emergency vehicle,” she said.

Rodak added that the collected data will then be processed with human involvement. “The human eye will be needed primarily for creating so-called annotations, i.e., marking various objects of the same type, such as cars, in the recordings.

“This will primarily be done by artificial intelligence, but it can make mistakes sometimes, and a human will be needed to verify its work,” she said.

The database is designed to train control algorithms for autonomous vehicles, helping them respond correctly in rare and complex traffic situations. Training will largely take place in computer simulations.

“Research institutions and companies will be able to use the scenarios we have created to train algorithms that control autonomous vehicles,” Rodak said, adding that the database will be non-profit and available to companies, universities, and researchers.

“For example, Blees, a company that already manufactures autonomous buses, operates in Gliwice. We also expect significant interest from the academic community,” she said.

The researchers also hope the database will attract international interest. “We would like our database to be used by well-known manufacturers. It would make it easier for them to introduce vehicles that will perform well in our conditions. Autonomous cars require training in specific infrastructure, signage, etc.,” Rodak said.

But she noted that fully autonomous vehicles are not expected on Polish roads in the immediate future.

“Although autonomous vehicle technology is developing, the challenges the engineers are facing have been quite surprising,” she said. “The confrontation with reality was devastating. For example, it turned out that there were no sufficiently precise sensors to equip fully autonomous cars. The traffic situations requiring algorithm training are incredibly numerous and highly complex. We cannot even catalogue all the situations that can occur on the road. There are far more of them than initially thought.”

Rodak said these challenges explain why fully autonomous vehicles do not yet exist. Currently, such cars operate under human supervision, such as the first certified Level 3 Mercedes-Benz system, or only within defined, safe areas, like Waymo’s robo-taxis.

“Thus, Waymo’s technology represents Level 4 automation according to the SAE J3016 standard. Such vehicles do not require driver intervention in some situations, but in others they do, and this is where the issue of defined operating conditions arises (e.g., types of traffic situations, specific weather and lighting conditions),” Rodak said.

But despite the limitations, Rodak she remains confident that fully autonomous vehicles will eventually be deployed on streets.

“It is difficult to make predictions. As scientists, we hope that this issue will be resolved within 10 years. I mean, we will be able to use cars that will drive completely independently. However, it is difficult to say anything about their cost at this time, and this will strongly impact their availability,” she said.

Marek Matacz (PAP)

tr. RL

mat/ agt/ lm/