Tallinn University of Technology develops new algorithm to make self-driving cars safer

According to foreign media reports, a research team at Tallinn University of Technology (TalTech) has developed a new algorithm that allows self-driving cars to drive smoother and faster. TalTech researchers led by Raivo Sell described simulation and real-life experiments, saying that the new algorithm makes self-driving cars "significantly smoother and more reliable", with a level of control close to that of skilled human drivers.

(Image source: Tallinn University of Technology)

Autonomous vehicles, including self-driving minibuses, are expected to be launched soon in many cities around the world. However, most of these projects are still in the experimental stage. TalTech said that these autonomous vehicles can meet the expectations of developers in a strictly controlled environment, but cannot cope with unexpected situations.

“One of the main strategies for autonomous vehicles is to avoid collisions with parked vehicles by changing lanes. This may seem like a simple task for humans, but it is difficult for most L4 autonomous vehicles on the market to accomplish this task. Manufacturers and related research are aiming to improve vehicle safety while reducing accidents and environmental impact.”

The first generation of self-driving cars can only operate in specific environments due to their simple algorithms. But new L4 and L5 self-driving cars are expected to be able to approach the basic skills of human drivers. This means that self-driving algorithms must be able to support increasingly complex actions, such as changing lanes, overtaking, and responding to unexpected situations.

According to TalTech, various data show that about one in ten accidents is related to the vehicle changing lanes in some way, which poses a greater challenge to autonomous vehicle developers. TalTech points out that "overtaking algorithms require comprehensive information about the environment in all directions and in all situations, as well as complex calculations about static and dynamic objects. When developing algorithms, other factors must also be considered, such as different weather conditions, traffic conditions and road quality, as well as communication with other road users, including vehicles, pedestrians and animals that may be running in front of the vehicle."

The new algorithm allows self-driving cars to make decisions about overtaking and changing lanes more quickly. "Although the new algorithm cannot yet reach the level of human drivers, it is still an important improvement," TalTech said. The university also said that this method can also help plan safe routes, using data proven by test results to clearly show which lane is safe. "In the new algorithm, the human-machine interface that shows the intention to overtake has been improved."

Raivo Sell said the new method has been tested in iseAuto, a self-driving car developed by TalTech. "We have collaborated with a research team from Aalto University in Finland to apply their theoretical machine learning model to our actual self-driving car to improve safety and add an algorithm that can abort overtaking when necessary to ensure safety."