Motion Recognition Enabled Through Single-Chip Processor Integration

The Electronics and Telecommunications Research Institute (ETRI) announced that it has developed a chip to process data collected from all sensors in autonomous vehicles.

It is a processor chip that executes image recognition and control functions required by autonomous vehicles with low power consumption around 1 watt (W).

This processor was named "Aldebaran" last year. It is one of the first-magnitude star designations. The core processor that determines autonomous vehicle performance had previously relied on overseas technology, but with domestication, there is no longer any need for overseas dependence.

In this research, following last year, the field where performance was improved is that processor cores were increased from 4 last year to 9. As the number of processing cores increased, processing speed accelerated, enabling clearer and larger image implementation.

Recognition capabilities also significantly improved. Currently, it supports real-time ultra-high-definition (UHD) image processing along with pedestrian, vehicle, lane, and motion recognition. Of course, radar and GPS signal processing recognition experiments were also successful. In the future, it is planned to be applied to Lidar and ultrasonic as well.

In particular, ETRI integrated the processor chip into a single chip (one-chip). The research team incorporated camera image processing functions and strengthened driver assistance systems to enable motion recognition on the chip.

Additionally, the chip also includes a black box function that can save and playback driving footage for vehicle security and accident evidence preservation. It supports UHD-level resolution compliant with the High Efficiency Video Coding (HEVC) standard.

ETRI, by fabricating it as a single chip (one-chip) to enable image processing, has been able to reduce the unit cost of the chip. The size of the chip developed by the research team is 7.8 x 6.7mm, smaller than a fingernail. The chip manufactured in this way is embedded in an electronic control unit (ECU) board (10cm x 10cm), passes through housing, and is embedded in the vehicle console area.

ETRI developed an autonomous vehicle-dedicated processor last year and transferred the technology to NextChip Co., Ltd., a semiconductor design specialist company. The newly developed chip is also planned to be transferred to relevant companies in the second half of this year, achieving mass production and commercialization next year.

ETRI explained that the performance of the Aldebaran chip developed this time is at the global level, and while the module prices of discrete chips offered by global competitors are in the hundreds of thousands of won range, these can be reduced to the tens of thousands of won range, demonstrating price competitiveness.

In the future, the research team plans to utilize neural network technology to incorporate high-performance artificial intelligence technology into the image recognition engine and achieve chip implementation.

ETRI's Kwon Young-soo, head of the Processor Research Group, also stated, "Our future goal is to accurately recognize all objects moving like humans. It is expected that chip development will be possible to enable services where machines can communicate with people, set destinations, and find routes on their own."