～Achieve high precision using the Quasi-Zenith Satellite ～Mitsubishi Electric's unique technology will
change the future of autonomous driving.
To realize autonomous driving, a high-definition locator（HDL） is needed for detection of accurate vehicle position information.
Therefore, Mitsubishi Electric has developed a HDL that achieves high accuracy by using the positioning augmentation data from the Quasi-Zenith Satellite.
In addition, we have succeeded in making high-definition map（HD Map）data, which would otherwise be enormous in size, compact with our unique compression technology.
The development team members who analyzed the key factors needed to realize this technology.
Car Multimedia System Engineering Dept.
Sanda works, Mitsubishi Electric CorporationTakaharu Eguchi
Development manager of HDL
Involved in engineering car multimedia products since 1993.
Is managing the technological development of HDL as the manager of the Car Multimedia System Engineering Dept. since 2019.
Car Multimedia System Engineering Dept.Yashushi Kodaka
In charge of designing and developing the map database for HDL.
Involved in the development of the map database for HDL since 2005. Since 2014, involved in developing high-definition map data for HDL.
Is developing high-definition map data and the map format using our unique compression technology in collaboration with map suppliers.
Car Multimedia System Engineering Dept.Masatoshi Fujii
In charge of designing and developing the position estimation function for HDL.
Involved in developing the position estimation function for the car navigation system since 2000. Since 2014, involved in the development of the high-precision position estimation function for HDL.
Mainly in charge of the development of the algorithm for the functions which collate the high-definition map with the location, and that link the location with the camera.
Car Multimedia System Engineering Dept.Tadatomi Ishigami
In charge of designing and developing the position estimation function for HDL.
Involved in the development of the position estimation function for the car navigation system since 1987. Since 2014, involved in development of the high-precision position estimation function for HDL.
Mainly in charge of the development of the GNSS high-precision positioning algorithm and the development of the composite positioning algorithm that utilizes dead reckoning.
※Profile is as of March, 2021
A high precious position is realized with high accuracy by the original positioning algorithm.
Since the setting angle is free, customers can attach this unit more easily.
FujiiThe positioning accuracy achieved of conventional car navigation systems is about 10 meters.
By using advanced positioning technology and positioning augmentation signals from the Quasi-Zenith Satellite※, we have improved the positioning accuracy to on average about 25 centimeters.
Furthermore, an error of 50 centimeters or less is realized with a probability of greater than 95% by the unique positioning algorithm.
With conventional car navigation systems, there are many restrictions on the installation angle of the device. However, we have made it possible to flexibly respond to any angle required with our unique altitude detection technology using high-precious positioning.
This technology makes it easier for customers to design the layout of their equipment inside their vehicle than ever before. ※The Quasi-Zenith Satellite System is a project led by the Cabinet Office, and Mitsubishi Electric is in charge of designing and manufacturing the satellite system.
Positioning accuracy / accuracy probability
Detects installation posture
Map accuracy improved from "road level" to "lane level"
A significant evolution toward using high-definition map data for autonomous driving
KodakaEven if there are multiple lanes on the road, only one line is used for description of the road.
By applying surveying technology, HD maps can depict the shape of each lane in detail with an error of less than half a meter.
This makes it possible to express in detail, for example, a branch road from its beginning to its end.
By combining the HD map with the high-precision positioning, the HD map can be used for decision making when advanced autonomous driving is being used.
As a result, on a motorway the system can control the speed of the car when approaching a curve or tollbooth.
Also, it is possible to assist lane changes based on information of the driving lanes and lane markings.
Feature of HDL
Maps can be used for autonomous driving as maps become more accurate
Vehicle position estimation function
Map provider function
Compressed the huge data size of the high-definition map
by making functions from a curved line
KodakaOriginally, the map data is high-definition data originally created from laser point cloud data acquired from surveying.
If the data is installed as it is, the map data will be about 6 GB for all the highways and motorways in Japan. Therefore, it was difficult to achieve sufficient performance with limited CPU and memory resources.
Also, there was the problem that the size was too large to update the map with a limited communication bandwidth.
Therefore, we reduced the number of shape points that make up the curve by using a curve-fitting method.
Of course, dissociation with the original map will occur. But there is an error in the original map itself, we thin out the map within the range of error that is acceptable for practical use.
As a result, the size of the HD map data has been reduced to about 40 MB.
Adopted a original algorithm that selects only high-quality signals to solve multipath effects
IshigamiThe accuracy of GNSS positioning is improved by using the positioning augmentation data from the Quasi-Zenith Satellite. However, the phenomenon of multipath, which is delay of radio waves made because of reflection by buildings, is inevitable due to the nature of radio waves.
Especially in a car where environment of radio wave reception is constantly changing it is difficult to perform accurate positioning under multipath, even if a positioning augmentation data is used, in situations under multipath conditions.
Far from accuracy in the centimeter class, it may cause an error of several meters. Therefore, we have developed an algorithm that checks each signal and selects only high-quality signals.
In addition, if the accuracy is reduced or the error is judged to be large, safety is ensured by considering the possibility of refraining from using it for advanced driving support.
FujiiIf the satellite positioning environment is unstable due to the multipath effect, GNSS positioning and dead reckoning is combined to ensure position accuracy.
After this is performed, the HD map is collated.
The combination of these technologies is a major feature of our HDL, and this makes it possible to accurately detect the driving lane even in snow where the road surface condition cannot be ascertained.
Specialized unit that estimates vehicle position and provides map data
in front of vehicle for AD/ADAS
Aiming for further accuracy improvements with
combined positioning using information from positioning and external sensors
EguchiWe have increased the accuracy on expressways from the level of 1.5 meters accuracy using just HDL to 0.5 meters using Quasi-Zenith Satellite and its positioning augmentation data.
In the future, we plan to achieve an accuracy of 0.5 meters even on surface roads by utilizing sensors such as cameras and road-to-vehicle communication using 5G communication technology.
Achieve 50 cm precision by utilizing the CLAS (Centimeter Level Augmentation Service) signals from the QZSS (Quasi-Zenith Satellite System)
HDL are a mandatory sensor for autonomous driving.
We will respond to the global use of HD maps with our advanced technology.
EguchiWe are receiving inquiries from various companies both in Japan and overseas about these developments.
Each of those companies is considering HDL as one of the sensors for autonomous driving level 2 or above.
In order to meet the needs of our customers, we have obtained advanced technical capabilities for various map companies in Japan and overseas.
KodakaIn Europe and the United States, there are services that deliver centimeter class positioning augmentation data for realizing 0.5 meter class accuracy. By using that information, high-precision positioning can be achieved.
Regarding HD maps, overseas map suppliers have been developing them mainly for highways.
By using them, it is possible to utilize HDL globally.
Using for MaaS
~ Providing data services that utilize high-precision location information to meet various social issues
and industry needs ~
EguchiWith the aim of providing data that enables local governments and road management companies to streamline and improve infrastructure maintenance and management, we are making efforts to add high-precision location information to image analysis results such as for road damage and deterioration.
This is just one example, but through these efforts, we aim to offer a "vehicle information data service" platform that enables analyzing and processing big data collected by various in-vehicle devices such as HDL on the cloud, and provide it to a wide range of industries.
By offering a platform focusing on use cases for MaaS, we aim to expand utilizing of our HDL to provide the safety, eco-friendliness, and operational efficiency of society as a whole.
Building Mobile Data Service Platform
Road Degradation Data Service
In 2021, a car that achieves autonomous driving level 3 has been released. This car also has a high-definition locator installed.
The importance of high-precision positioning is expected to increase, but there are still many challenges.
To solve these problems, Mitsubishi Electric produced HDL by utilizing our positioning technology and algorithms and our unique compression technology for using HD maps.
We aim to provide not only a platform for autonomous driving but also vehicle information data services to solve various social problems and meet industry needs.
Remarkable points of Mitsubishi Electric HDL
- It achieves highly accurate positioning with high probability by using a unique positioning algorithm.
- The data size can be reduced by using the data compression technology. It brings advanced technology to a practical level.
- It addresses multipath effect in combination with self-sensing driving.