What are the differences between in-vehicle high-precision maps and traditional maps?

At present, autonomous driving has become one of the important directions of development in the field of artificial intelligence, and high-precision maps are often called the "clairvoyance" of autonomous driving. So what are the differences between high-precision maps and traditional maps?

"Compared with traditional navigation maps, high-precision maps must not only accurately recognize lane models, road components, road attributes, etc., but also clearly judge dynamic traffic information such as road congestion, construction conditions, and whether there are traffic accidents." Liu Jun, founder and CEO of Kuandeng Technology, showed reporters the latest research and development results. The high-precision map in front of him not only presents the road shape, slope, curvature, paving, direction and other information of the basic map, but also has detailed data annotations on lane line type, lane width, roadside landmarks, guardrails, road edge type and other aspects.

Rapid information collection in complex scenarios Real-time update of road dynamics

It took only 24 hours from the announcement of the opening of the Hong Kong-Zhuhai-Macao Bridge to the production of a high-precision map data. Rapid and accurate data collection is the first step in completing a high-precision map.

"Without collected data, the machine can do nothing. High-precision maps rely on collected data to be produced. The latest collection software can help collectors solve many problems." Zou Yang, senior product manager of Kuandeng Technology, introduced that the current collection equipment is more flexible and rapid in technology development and iteration, and has stronger capabilities for large-scale expansion configuration and rapid collection and production, and faster map data updates. Even data collection in bad weather conditions is not difficult.

What if the scene is dark and you can't see anything, such as in a tunnel?

"When faced with complex overpasses and extra-long tunnels without GPS signals, our automation technology can still automatically generate infrastructure such as street lights and signboards, road signs, and even dotted lines on the road. With such rich road information, autonomous driving vehicles can also drive safely in the Qinling Zhongnanshan Tunnel, which is known as the 'No. 1 Tunnel in the World,'" Zou Yang introduced.

But for high-precision maps, the completion of a map drawing and delivery does not mean the end, but the beginning of a closed loop of map updates. "If a high-precision map cannot guarantee continuous updates, the security risks caused by information lags will continue to increase. The completion of a high-precision map requires field staff to collect data, which is then further processed by office staff and finally updated to the map through a series of algorithms." Zou Yang said that in the face of ever-changing road information, in terms of map content updates, Kuandeng Technology uses a method of mutual verification between different intelligent information sources to improve credibility and ensure the continuous update of high-precision maps.

AI technology generates knowledge data, every blade of grass and every tree is clearly visible

Nowadays, many companies have entered the field of high-precision maps, including many Internet giants, but in this emerging battlefield, core technology is the key to victory. One of the core technologies of high-precision maps is to make full use of artificial intelligence technology in the automated processing process. Through deep learning and model training of different urban scenes and different road scenes, a large amount of knowledge data is formed, and the artificial intelligence system automatically completes the recognition, feature point extraction, precise measurement and lane network topology construction.

Sun Xu, product director of Kuandeng Technology, showed reporters a large number of high-precision maps of different scenes in cities such as Beijing, Shanghai, Guangzhou, etc. In these high-precision maps, the precise location information and semantic information of various features on the road, including lane lines, ground markings, traffic signs and other road facilities, are fully and accurately displayed in the Kuandeng high-precision maps.

In the high-precision maps of several overpasses on the East Fifth Ring Road in Beijing, the shape of the entire road can be seen, including every lane line on the road, and even the starting and ending points of each lane line are expressed.

"Interchanges are usually very challenging places in the research of autonomous driving technology, because an overpass usually has more than a dozen road branching points, such as road bifurcations, main road ramps, ramps into main roads, etc. Each overpass is very complex, and each analysis point requires us to collect data multiple times." Sun Xu said that through AI's automated processing technology, in the high-precision map on display, you can see that the map not only expresses the shape of the road, but also each lane line and the position of each branch starting point are accurately expressed, so that autonomous vehicles can get accurate decision-making information when there are vehicle changes, road changes, and lane following.