New Japanese technology helps autonomous vehicles recognize traffic lights 150 meters away

2024-02-29

With the continuous development of autonomous vehicle, such vehicles must be equipped with systems that can work in coordination with existing infrastructure, and need to be able to handle complex urban landscapes. For example, handling traffic lights of different shapes, sizes, and installation positions; Multiple city logo colors; Arrow lights indicating direction. In order to make the autonomous vehicle run safely, it must be able to recognize such objects in a short time, understand them and respond accordingly.

Therefore, according to foreign media reports, researchers at Kanazawa University in Japan have developed a new method that can reliably assist autonomous vehicles in handling complex traffic conditions and different arrows indicating directions in cities. This system allows vehicles to recognize traffic lights 150 meters away and smoothly and naturally slow down or adjust driving operations.

In urban areas where vehicles, pedestrians, and various objects are densely packed, autonomous vehicles face special challenges, typically requiring the use of sensors to estimate position and high-precision digital maps that can provide precise positions of stationary objects to handle this situation, as these technologies can help vehicles estimate their own position and surrounding conditions, thereby more intelligently changing position or speed. In all situations, successfully dealing with traffic lights is crucial.

Ryo Yanase, co-author of the study, said: "In order for autonomous vehicle to respond safely, they must be able to see traffic lights at least 100 meters away. Therefore, we have created a unique algorithm that uses digital maps to identify relevant traffic lights, and then calculate the necessary acceleration or deceleration operations. We have also found a way to make autonomous vehicle better than most people at identifying arrow lights in the distance (indicating direction). "

Using a high-resolution lens or telephoto lens to recognize traffic signals from 100 meters away may seem like a relatively simple task, but when quick decision-making and a wide field of view become crucial, processing time may increase and the field of view may be reduced. The algorithm solution based on digital maps developed by researchers, in collaboration with sensors, can draw the position of vehicles and the "area of interest" containing traffic lights when approaching intersections.

The system does not have too many resources, so the central processing unit (CPU) of the vehicle can perform real-time processing. Researchers will apply this method to various urban driving conditions to see how it works. It was found that even if separated by 150 meters, the system can easily recognize traffic signals and adjust the speed accordingly.

The first author of the study, Keisuke Yoneda, stated, "We have successfully created and validated two valuable achievements. Firstly, we are able to detect traffic signals at long distances and respond with almost no false positive detection, and can still manage CPU load. On the other hand, our system can recognize small arrow like objects, even objects with pixels less than 10."

The above research results will contribute to the development of necessary technologies to enable autonomous vehicles to travel more safely in busy urban areas filled with other vehicles, pedestrians, and many objects.

Simulation technology unleashes the potential of 3D printing additive manufacturing