Dynamic Map to generate 3D maps for Japan's expressways by 2018

Promoting plans for standardization of specifications, efficient production, and map generation

2017/02/14

Summary

Dynamic Map Corporation will utilize the the Mitsubishi Electric Corporation developed Mobile Mapping System (a mobile high precision 3D information platform). Mitsubishi Electric is one of the company's shareholders (Source: Eisan Technology Co. Ltd.).
Dynamic Map Corporation will utilize the Mitsubishi Electric Corporation developed Mobile Mapping System (a mobile high precision 3D information platform). Mitsubishi Electric is one of the company's shareholders (Source: Aisan Technology Co. Ltd.).

 The Automotive World 2017 conference was held at Tokyo Big Sight from January 18 - 20, 2017. This report features an overview of a lecture given by Mr. Masamichi Hamada, Vice President of Dynamic Map Infrastructure Planning Co., Ltd. (hereafter referred to as Dynamic Map Corporation) titled, "Put "Dynamic Map" into practice and contribute to automated driving."

 Dynamic map are a 3D map that combine dynamic information such as traffic jams, accidents, vehicles in the vicinity of an automobile with static information about structures like roads and buildings, and are indispensable for realizing autonomous driving. Dynamic Map Corporation was established as a planning company to examine data specifications for these 3D maps. The company’s work is ongoing, and it is also looking into generating 3D maps itself.

 To generate dynamic maps, point cloud data is measured by a LIDAR equipped vehicle as it travels on the target road. Images taken with a camera at the same time are also utilized to generate "vector data," which is easier to use. For 3D maps, the major challenges are finding methods to efficiently generate vector data and update it so it reflects the latest situation as the conditions of roads change every day.

 The Japanese Government is also promoting 3D maps as part of the Strategic Innovation Promotion Program (SIP) led by the Cabinet Office and aims to realize dynamic map generation for highways and expressways in Japan by 2018.

 Moreover, the company has already started negotiations with organizations in other countries in order to ensure that the dynamic maps it creates do not have standards that differ significantly with similar systems made overseas.

Related report:
3D Maps for Autonomous Driving: Standardization and Update Method Studies Advance(September 2016)



Dynamic Map Corporation founded with an aim for specification standardization

 Dynamic Map Planning Co., Ltd. was established in June 2016 to plan data specifications and maintenance methods commonly used by automobile manufacturers and map companies in cooperative areas of dynamic maps necessary for autonomous driving. Mitsubishi Electric, which developed a mobile mapping system (mobile high precision 3D measuring system), map companies, survey companies, and 9 automobile manufacturers have invested.

 Dynamic Map Corporation is also considering taking the lead in generating 3D maps from actual measurements.

 (Note)  The creation of 3D maps for autonomous driving will be divided into cooperative and competitive areas. Cooperative areas will call for the various map makers and OEMs to standardize specifications, even as the companies seek to differentiate themselves in the competitive areas. Dynamic Map Corporation will handle specification standardization in cooperative areas.



The role dynamic maps play in autonomous driving

 Advanced "self-positioning estimation" and "peripheral environment recognition" are necessary for autonomous driving systems. Although self-positioning estimation is usually performed with a combination of a Global Navigation Satellite System (GNSS) such as GPS and a sensor, as these are subject to accuracy problems and are dependent on climatic conditions, highly accurate digital maps are necessary for autonomous driving.

 (Note) GNSS is a generic name for satellite positioning systems such as GPS, GLONASS, Galileo, and the Quasi-Zenith Satellite System (QZSS).



Generating vector data from point cloud data with LiDAR and adding virtual data

 Dynamic map will use MMS to acquire point cloud data (data in the form of a large number of points with 3D positioning information) (see the images to the bottom left and center), and since it would be difficult to use the data as is, vector data (image to the bottom right) is generated through calculations based on numerical data including coordinate points and the lines that connect them, along with camera images taken simultaneously by MMS. The vector data has length, directionality, positioning, and attribute (information such as "this line is a road" and "this data represents a car."), and this enables verification and recognition of features such as lanes, traffic lights, and stop lights (everything on the ground is called a "feature" in surveying terminology). Vector data is also characterized by having smaller file sizes than point cloud data.

Generating vector data

 Dynamic Map Corporation has studied the procedures defined in previous research and surveys about the attributes of features that are difficult to identify with the MMS point cloud data. The company aims to identify more features by utilizing AI in the future.

Adding virtual data

 Dynamic maps also display virtual objects to represent other information about lanes useful for autonomous driving. These include virtual lines in the center of the lanes and lane links (lines are extended and connect to others at intersections).

Laser point cloud data acquired with LiDAR. Dynamic Map Company generates vector data from here by utilizing camera images filmed simultaneously and other information. (Source: Aisan Technology CO.,LTD) Point group data of the Ohashi junction measured by MMS. The data was measured by a surveying vehicle as it traveled through the tunnel, and this allows it to be output as a downward view image like this. (Source: Pasco Corporation) A vector data image. It displays features such as utility poles/signs, traffic signals, and road edges. Virtual feature such as a lane links (virtual indicators useful for automated driving) are displayed in purple. (Source: Aisan Technology CO.,LTD).
Laser point cloud data acquired with LiDAR. Dynamic Map Company generates vector data from here by utilizing camera images filmed simultaneously and other information. (Source: Aisan Technology CO.,LTD) Point cloud data of the Ohashi junction measured by MMS. The data was measured by a surveying vehicle as it traveled through the tunnel, and this allows it to be output as a downward view image like this. (Source: Pasco Corporation) A vector data image. It displays features such as utility poles/signs, traffic signals, and road edges. Virtual feature such as a lane links (virtual indicators useful for automated driving) are displayed in purple. (Source: Aisan Technology CO., LTD).



Constructing a dynamic map with information divided into 4 levels according to time

 Dynamic maps are generated by dividing map information based on time into 4 stages and collecting it as shown in the figure on the right. The level at the very bottom is itself a highly accurate 3D map. It shows "static information (acquired on a monthly basis)" that does not change easily such as road surface information, lane information, and structures. The levels above handle information that changes constantly in a short period in increasing detail. The top "dynamic information (grasped every second)" level includes surrounding vehicles, pedestrians, and ITS prefetch information that Toyota has put into partial practical use. The data in these levels is gathered with location information occupying a central role. This scheme is characterized by the central management of all information necessary for autonomous driving.

 The "static information" at the bottom and "semi-static information" level above it are the cooperative areas that Dynamic Map Corporation works on. The image shows a model originally proposed by a European planning organization and is called a "local dynamic map."

ダイナミックマップの内容を時間軸により4つの段階に分類した。この4枚を、位置情報をキーに重ね合わせてダイナミックマップを生成する。(資料提供・ダイナミックマップ社)
Dynamic map contents are classified into 4 levels based on the time. These 4 images are superimposed on the positioning information to generate a dynamic map. (Source: Dynamic Map Corporation)



Cooperative and competitive areas of dynamic maps

 As part of the Strategic Innovation Promotion Program (SIP) led by the Cabinet Office, the Japanese Government is supporting the creation of dynamic maps. In 2015, Dynamic Map Corporation shareholders such as Mitsubishi Electric and Aisan Technology Co., Ltd have launched various research programs looking into the specifications of 3D maps for autonomous driving as part of the SIP program.

Distinction between cooperative and competitive domains

 A summary of the result of these inquiries can be seen in image to the bottom left. Dynamic Map Corporation is in charge of the standardization of specifications for map information (the area bordered by the thick red line) corresponding to the cooperative area of "recognition," which is one of 3 actions normally performed by the driver (recognition, judgment, and operation). The areas enclosed by dotted orange lines are cooperative areas not overseen by Dynamic Map Corporation, but are instead overseen by the Cabinet Office's SIP initiative.  ”Judgment" and "operation," which are highlighted in blue, are competitive areas where map companies and OEMs seek to differentiate themselves.

Putting the final touches on dynamic maps by combining data from competitive and cooperative areas

 The figure on the bottom right shows an outline of the procedure for creating a dynamic map, as well as areas Dynamic Map Corporation needs to investigate. Common base data is collected from laser point cloud data and video. This data is used to generate static information vector data and construct a "high precision 3D positioning information platform" (the area bordered by the thick, light brown line is of the area Dynamic Map Corporation works in).The company then completes the map by taking the following into account: public information, semi-static information, semi-dynamic information, and dynamic information. It is currently discussing areas including specific procedures to finalize the details as soon as possible.

 Based on this cooperative area data, map companies and surveying companies will add their own data (from competitive areas) to complete their map and then provides it to automobile manufacturers. This policy also calls for developing applications for public enterprises and other uses in addition to autonomous driving.

Collaborative and competitive areas for dynamic maps. The area bordered by the thick red line indicates the areas covered by Dynamic Map Corporation (Source: Dynamic Map Corporation)
Collaborative and competitive areas for dynamic maps. The area bordered by the thick red line indicates the areas covered by Dynamic Map Corporation (Source: Dynamic Map Corporation)
The work flow from information collection to provision of dynamic maps. The area bordered by the light brown line indicates the areas covered by Dynamic Map Corporation (Source: Dynamic Map Corporation)
The work flow from information collection to provision of dynamic maps. The area bordered by the light brown line indicates the areas covered by Dynamic Map Corporation (Source: Dynamic Map Corporation)



Future Challenges: Advancing the timetable for practical use, improving efficiency in maintenance, and other difficulties

Progress in improving maintenance efficiency (Source: Mitsubishi Electric Corporation)
Progress in improving maintenance efficiency (Source: Mitsubishi Electric Corporation)

 Dynamic Map Corporation highlighted the following 4 items as future challenges:

  1. Advancing the timetable for practical realization;
  2. Improving maintenance efficiency;
  3. Global compatibility;
  4. Expanding applications.

 (1) To advance the timetable for practical realization of this technology, the government aims to generate 3D maps for expressways and highways in Japan by 2018. It is necessary to generate vector data and add virtual feature data when generating dynamic maps. Even after surveys have been performed with MMS equipped vehicles, various tasks still need to be performed to create dynamic maps including the generation of vector data and addition of virtual feature data. Dynamic Map Corporation is looking into means to improve efficiency for these processes.

 (2) Improving maintenance efficiency" is also an urgent issue. Since road conditions will be constantly changing even after 3D maps are completed, procedures are being investigated for having survey vehicles equipped with compact LiDAR update measurements when a difference with existing data is found.

 The figure on the right is an example of a case where maintenance efficiency improvement is progressing. The images on the left show a difference being identified in previously measured data (the picture on the upper left) and corrected with new data (the picture on the lower right). The image on the right shows the acceleration of point group processing .As the maintenance vehicle travels, it updates the people indicated in red, the boxes placed, and arrows in real time.

 (3) In the area of global compatibility, the company is investigating data specifications including for competitive areas that are easy to process in similar systems utilized overseas to prevent the development of different standards. It has also started negotiations with organizations in other countries that are working on such systems.

 (4) As for the expansion of applications, the company believes that there is a possibility of contributing not only to automated driving but also many other applications; including disaster situation visualization for disaster-resistant city development; pre- and post-disaster analysis as one element of improving social infrastructure management efficiency; maintenance and management of roads and rivers; and measures to deal with the aging population.

Keyword

3D map, Dynamic map, Dynamic Map Planning Co., Ltd., LiDAR

<Automotive Industry Portal MarkLines>