Raon Road, a subsidiary of Raon People (CEO Seok-joong Lee), announced that it’s the only and first in Korea to construct an intelligent transportation system (ITS) based on ‘artificial intelligence (AI) digital twin’. It is anticipated to be accomplished as early as next yr.

Kang Byeong-gi, Vice President of Raon Road, said on the twelfth, “By combining digital twin with 14 years of vision AI technology, we’re developing technology that goes beyond cameras and becomes 'eyes' that view all the city.”

Raon Road’s ITS is basically divided into two types. These are the Traffic Evaluation System (TAS), which analyzes intersection images using AI, and the Traffic Management System (TMS), which selects and displays only the vital information throughout the traffic center.

The core technology that makes this possible is ‘Vision AI’. It should be possible to quickly analyze objects by screens captured and transmitted from devices equivalent to CCTV. It must even be possible to detect statistics equivalent to the placement, speed, direction, and traffic volume (density) of vehicles and pedestrians, in addition to unexpected events equivalent to signal violations.

He said that ‘constructing an interface’ can be vital. That is to make it easier to utilize and understand video data beyond simply collecting it with a camera. The reason is that it’s a system that displays real-time object information and intersection information on a map, making it easy to research traffic data.

Vice President Kang Byeong-gi emphasized, “In these two points, Laon People has differentiated technology.”

This might be explained by the incontrovertible fact that its parent company, Laon People, is a listed company representing the primary generation of vision AI, established in 2010. Laon People has been researching the mixture of machine vision and AI even before the arrival of AlphaGo. The related knowledge posted on the SNS operated on the time was also used as a textbook amongst AI researchers.

As well as, Raon Road confirmed its spin-off in November 2021 and started to proceed its transportation-specialized AI business. Particularly, the national-level ‘Smart Intersection Project’ in 2019 served as a possibility. A wise intersection is a system that uses AI to gather and analyze various information, equivalent to intersection vehicle traffic volume and risk aspects, and improve traffic conditions.

It was revealed that through this process, they’d amassed know-how in constructing infrastructure and interfaces. Currently, technological advancement is in progress, and digital twins have been chosen because the core.

Digital twin is a kind of 'reproduction and prediction', and its principal goal is testing and inspection before actual problems occur. The goal is to copy the actual road structure and traffic conditions within the virtual world.

Through this, testing might be conducted before expanding roads or installing traffic lights. We predict how the flow of cars and traffic volume will change based on existing data.

He explained that it’s also essential when it comes to connectivity. Currently, each camera only records individual screens, but using digital twin, data between devices might be connected and displayed.

It’s identified that digital twins are also vital for efficiency, speed, and convenience. Nearly all of data collected by CCTV is unrefined ‘raw data’. It’s obscure not only subtle changes that should not visible on the screen, but in addition the general flow. There are also limits to the necessity for people to tackle the roles of managers and supervisors.

Nonetheless, if that is symbolized and standardized and presented in a clean interface, not only managers but in addition AI can intuitively understand it. It’s because learning and problem situation recognition develop into easier with organized data. By excluding unnecessary elements contained within the camera, data processing and capability might be reduced.

Currently, it has been reported that vehicle data has been collected and partially implemented as a digital twin in real time (inside 4 to five seconds of error). Even the estimation stage is feasible.

The ultimate step is technology to ‘match’ vehicle data on multiple roads. It goes beyond recognizing each object and predicts the flow. That is the stage where “it goes beyond the camera and becomes an eye fixed that appears at all the city.”

The explosive growth of knowledge shall be resolved through the cloud. That is to simply connect and share data beyond the constraints of existing on-premise. It has recently acquired Tdig, a cloud specialist company, to strengthen its capabilities.

There are also plans to supply cloud services in order that transportation solutions might be conveniently utilized in the long run. This has the advantage of collecting more data and strengthening AI.

They said in addition they plan to introduce generative AI and multimodal functions. A traffic evaluation system that permits questions and answers in natural language by applying large language models (LLM) and chatbots, and a multi-modal model that may explain CCTV footage in image units are also scheduled to be released by the tip of this yr or next yr.

Edge AI can’t be unnoticed. For pedestrian safety, AI cameras might be installed at each crosswalk to acknowledge objects and control and extend signals as needed. We’re collaborating with semiconductor specialists Mobilint, Rise up, and Furiosa AI.

Because of this, if Raon Road technology is applied to the road network, large-scale digital transformation (DX) equivalent to smart cities will develop into possible. Smart CCTV and smart intersections have already been installed in Seongnam, Gwangyang, Namyangju, and Andong.

Lastly, Vice President Kang Byeong-ki said, “As we now have been constructing a transportation AI business because the starting of the national project, we’re ahead when it comes to relevant data and problem recognition,” and added, “We will even be liable for core technology within the upcoming autonomous driving era.”

He also emphasized, “We’ll create practical AI with evaluation technology, which is the strength of large-scale AI, on the core.”

Reporter Jang Se-min semim99@aitimes.com