Preface: Spring recruiting season again! As a national travel service platform, AutoNavi’s business is developing rapidly, and a large number of school/social recruitment places are open. Welcome to submit your resume. See details at the end of this article. To help you know more about AutoNavi technology, we have planned
# Spring Recruit ColumnA series of articles organized by senior students of various business teams
Business science popularization + technology application practiceFor the main content for everyone to do the relevant introduction.

This article is the second of the “Spring Recruitment” column series. It is compiled according to the content of “High-precision Maps and Collection and Production System for Autonomous Driving” shared by Xiang Zhe, General Manager of AutoNavi High-precision Maps Business, AT the AT Technology Forum. Abbreviate the content without prejudice to the original intention.

At Technology Tribune is a technical exchange activity initiated by AutoNavi. Each issue will focus on one topic, and we will invite experts inside and outside Ali Group to give speeches, QA, and open discussion to make technical exchanges with you.

Xiang Zhe mainly shared two aspects of the content:

1. What is the high-precision map for autonomous driving;

2. Current construction status and thinking of high-precision map data acquisition and production system.

High-precision map is an indispensable core condition for autonomous driving vehicles, which needs to accurately express the spatial positions and relative relations of various elements in the real world. Therefore, the production of high-precision map requires high precision of data collection.

Industrial level autonomous driving classification

Combining with the application status of automatic driving in industrial level, he began to talk.

At present, industrial-grade autonomous driving can be roughly divided into two categories. The first category is represented by intelligent cars with “auxiliary driving function” produced for users by new car manufacturing forces such as Tesla and Xiaopeng. When using these automatic driving functions, the user should take over the driving at any time, and the legal problems arising in the driving process are mainly the responsibility of human beings.

AutoNavi has deep cooperation with the mainstream new forces in the market in the field of high-precision maps. Xiaopeng, for example, uses AutoNavi’s highly refined mapping capabilities for its driving assistance. What kind of assisted driving capabilities are currently available to users? It has basically realized the automatic driving function from point to point on the expressway.

For example, driving from Beijing to Guangzhou on the expressway will encounter several sections of expressway switching, that is, from one section of expressway to another section of expressway on the ramp, as well as lane changes to overtake on the expressway. The NGP auxiliary driving capability of Xiaopeng has the above two capabilities, and basically has the full self-driving function from Beijing to Guangzhou.

But in the process of driving, the driver should keep an eye on the road conditions at any time, in the automatic lane change if there is a risk of scraping to manually take over driving, continue to manually drive to complete the lane change. This means that the driver is always ready to take over while driving.

In addition, drivers will have to manually take over the driving of the car when passing through toll booths, as the current high-precision maps do not provide lane-level information in toll booths. When approaching a tollbooth, the voice assistant will inform the driver that there is no high-precision map on the road ahead of the tollbooth and that the driver needs to take over manually. The above autonomous driving capabilities will be applied to all of Xiaopeng’s P7 models.

The second industrial-grade type of autonomous driving is typical of the L4. For example, Google is making taxis and logistics trunk trucks that drive themselves in cities. Compared with the first type of automatic driving, this type of L4 automatic driving theoretically has no driver on the car. Although now in the verification phase, there are still drivers in the driver’s seat. Xiang predicted that it would take four to five years for such L4 driverless taxis to enter the lives of ordinary people.

Both types of autonomous driving rely heavily on highly refined maps.

Sophisticated maps and autonomous driving

High-precision maps are the maps in the self-driving car’s “head” that let it know what “invisible” road conditions will look like next. The four key functions of autonomous driving are perception, high precision positioning, decision planning, and vehicle control. There are at least three features that rely heavily on highly refined maps.

Perception: When driving a car, humans should observe the surrounding lane lines, traffic signs, poles and other information. Sensors in the smart car will pick up information about objects around the road. The highly refined map provides a super-visual perception of God’s perspective. In particular, when there is a large truck in front of the vehicle, human eyes and sensors can not see the lane line and other information in front of the vehicle, the high-precision map data can inform the vehicle of the road information ahead.

High-precision positioning: For an autonomous car to know exactly where the car is on a map, the premise is to rely on the base map provided by the high-precision map. Automated driving vehicles need to know their position on the map based on two capabilities. One is the absolute position information provided by GPS, inertial navigation, and Qianxin positioning capabilities. The absolute position information is matched with the latitude and longitude coordinates of the map, and the specific position of the vehicle on the map can be determined (depending on the ability of the sensor to locate the absolute position).

But only absolute positioning is not enough, in a particular area, such as high-rise buildings, canyons, etc. Will keep out signal occurs, absolute positioning accuracy will become worse, lane line around the autopilot rely on observation, signal, the rod of relative positioning to assist, to high precision of map data matching judgment. In actual projects, AutoNavi conducts in-depth cooperation with mainstream automakers to determine which technologies can achieve more accurate relative positioning capabilities.

Decision planning: Automated driving is highly dependent on road elements such as lane lines, traffic restriction facilities, and traffic lights to comply with driving rules.

The above functions support each other.

High-precision maps for autonomous driving

Several key elements: road layer, lane layer, location object, dynamic layer.

Road layer: HD and SD data are closely matched. At present, almost all automatic driving systems are first told by the user that I want to go from somewhere to somewhere, and the driving route planning between these two locations is supported by SD road data. HD data is not isolated, and SD data to connect. SD data capability is AutoNavi’s traditional strength, coupled with the industry leading HD capabilities, this match AutoNavi must be the best in the industry. This is also what car manufacturers look for when choosing a mapping service.

Lane layer: all of the autonomous driving ground level control of the vehicle relies on highly refined map data.

Positioning objects: AutoNavi works closely with the car factory to determine the relative positioning based on which technologies, which reference objects to choose, and what level of precision to achieve, etc. The two sides will communicate and conduct joint research and development together.

Dynamic Layer: Future high-precision maps will contain dynamic layers, real-time data, and what dynamic traffic events happen in a certain lane at a certain moment.

High-precision map in the city of ordinary road challenges

At present, AutoNavi’s high-precision map has completed the collection of more than 300,000 kilometers of highway and urban fast sections, and is entering a stable and regularly updated state. Compared with the expressway city fast, more difficult problem in the city ordinary road.

One of the key challenges to high-precision urban mapping data is intersections, many of which lack traffic lines on the ground. Autonomous cars that turn between junctions and drive across the ground with no traffic lines (paint) rely on highly sophisticated maps that have been prepared in advance. Of course, it is not only the ground lines at intersections that need to be considered, but also a large number of other traffic elements. But autonomous driving on the streets of cities is certainly an important scenario in which map providers and new car makers will devote a lot of attention in the future.

High precision map collection and generation

Conventional high-precision map production can be summarized into three stages: “collection”, “production” and “productization”.

The acquisition vehicle is a mobile acquisition system which is precisely integrated by a variety of advanced measurement sensors. Generally, it contains LIDAR, inertial navigation, camera and other equipment, and is equipped with different types of sensor equipment according to different acquisition scenes. After years of deep cultivation, AutoNavi High-precision team has developed the high-precision acquisition vehicle system, which has the characteristics of high precision, fast speed, short data generation cycle, high degree of automation, high security and information uniformity.

After the acquisition equipment collects the accurate data in the external real world, it has to go through the steps of image recognition, precision processing and manual processing to “become” the high-precision map data that can be used.

“Fresh” high-precision map

The road data in real life is constantly changing, how to achieve “fresh” high-precision map.

First, it will start with a relatively expensive, specialized underpinning vehicle that will measure and collect highly refined map data across the country’s roads. Both relative and absolute accuracy should be guaranteed. Then, relatively inexpensive professional renewal vehicles are used to collect local (relative) changes in road information, such as repainted ground signs, newly erected signs, poles, etc. At the same time, we also use cheaper crowdsourcing equipment to do faster collection updates.

In order to realize the rapid update of the existing data and improve the freshness of the data. AutoNavi’s high-precision team has built a collection system with three levels of capabilities, namely, professional base car, professional update car and crowd-sourcing update, to solve the problems of precision and freshness together. In real business scenarios, you need to find a balance between precision and freshness, and iterate repeatedly.

In order to solve the challenges of high precision map data, such as “high precision”, “large scale” and “fresh enough”, it is necessary to make breakthroughs in many technical points.

For example, how to design and manufacture acquisition and mapping equipment with different costs, different precision and different deployment capabilities;

How to cooperate with different types of equipment for collection, while meeting the requirements of precision and freshness;

How to design and apply algorithms to improve the absolute accuracy and relative accuracy of data collection, and ensure the alignment of multiple data collection;

How to comprehensively use image and point cloud to do a good job of identification and improve the automation level of production?

The students in the map team of AutoNavi focus on different directions, accept business challenges with an open mind, discuss and design solutions together, and have made a lot of achievements.

With this high-precision map with the highest accuracy and the most extensive coverage in China, AutoNavi has successfully won commercial orders from many mainstream car manufacturers at home and abroad, and started to provide high-precision positioning, over-the-horizon perception, lane-level navigation and other services for intelligent driving models. As AutoNavi’s key breakthrough field in the autonomous driving ecosystem, AutoNavi’s high-precision map business is developing fast and has many opportunities. We hope you can join us.

About the high precision map business center

High-precision mapping is one of AutoNavi’s most innovative businesses, dedicated to measuring the world with sensors, understanding the world with algorithms, and redefining the world with data. We cover almost the most popular frontier disciplines, high-precision mapping and autonomous driving are multi-disciplinary application engineering systems. Based on perceptual understanding, 3D reconstruction, fusion positioning and computational geometry technology, high precision digital 3D map is automatically generated. Using edge computing, big data processing, cloud services, to carry out real-time map reconstruction of massive data. Through 5G/V2X information exchange, data exchange between map objects is realized to build a live map. We’re not just data makers, we’re life definers. Join us and the future is “up to” you.