Initiatives to Achieve Our Goal of Zero Fatal Traffic Accidents in 2030

The stereo cameras equipped in EyeSight enable the system to recognize objects with two cameras, similar to human eyes, detecting items on the road at a higher level than radar-based systems. The development of EyeSight began approximately 30 years ago. During the early stages, there were frequent issues with the system’s ability to detect objects properly, particularly when it was raining or the windows were foggy. However, these challenges were overcome through extensive testing in various real-world scenarios and subsequent improvements.

The EyeSight system was launched in 2008, achieving a world first in using solely stereo camera technology to enable pre-crash braking and adaptive cruise control. EyeSight version 2 was announced in 2010 with an enhanced pre-crash braking system that provides support until the vehicle comes to a full stop. During this period, EyeSight became a standard feature in numerous models for the Japanese market, and its affordability contributed to a significant increase in its recognition and widespread adoption among customers. In addition, it began to be developed in certain models overseas as well.

In 2014, with the release of version 3, advancements in color recognition and an expanded field of view using stereo cameras significantly enhanced the advanced safety features. In the 2020 release of the Levorg, Subaru introduced EyeSight X, combining a new stereo camera with four radars in the front and rear, and a highprecision map locator.

The fundamental principle of spatial recognition using stereo cameras is to precisely calculate the distance to objects from each pixel by triangulating the displacement of images captured by the two cameras, one on the right and one on the left. The strength of EyeSight lies in its ability to accurately convert everything captured by the cameras into a three-dimensional representation, allowing it to perceive the shape and distance of various objects. However, when it comes to capturing small objects or subtle irregularities, determining whether to pass over them or consider them obstacles solely based on EyeSight‘s image recognition can be challenging. In contrast, AI excels at precisely categorizing objects captured in camera, pixel by pixel, based on extensive previously learned data. Since EyeSight can fuse the distances to recognized objects with the classifications made by AI in the same image, EyeSight and AI are considered an excellent match. If this can be put into practical use, it is expected to lead to an enhanced level of vehicle safety in various situations by allowing for more detailed and accurate recognition of the driving environment’s information than ever before.

Subaru focuses on developing cars with the utmost priority of safeguarding everyone in the event of a collision, not just the vehicle occupants but also pedestrians and cyclists.

Strengthening the vehicle body and enhancing restraint systems are two critical elements to occupant protection. In strengthening the vehicle body, we combined a cabin structure that can withstand severe collisions by using high-strength materials with the Subaru Global Platform, which helped realize a body structure that efficiently absorbs collision energy. This design ensures protection in collisions from any direction.

Pedestrians and cyclists sustain injuries at a rate 2.5 times higher than that of vehicle occupants. To reduce the damage to pedestrians and cyclists in the event of an accident, Subaru not only uses soft structures for bumpers, bonnets, and other areas that can absorb impacts but also equips its vehicles with pedestrian airbags.

In the pursuit of zero fatal traffic accidents, it is necessary to anticipate more diverse and severe accidents than ever before, however. For example, in collisions involving bicycles, it has been recognized that under certain limited conditions, the current pedestrian airbags may not cover all cases, given the various factors such as the cyclist’s physique and the orientation and speed of both the bicycle and the vehicle at the time of impact. As a result, we are utilizing computer simulations to identify worst-case scenarios from the countless collision patterns that can be imagined and to develop specific countermeasures.