Introduction:From the first steam driven three wheel vehicle in the world produced by the Frenchman, N. J. Cunnough, in 1769, to the modern automobile with the development trend of automation, electrification, networking and sharing, science and technology has always been the cornerstone to promote the development of the automobile industry. In recent years, with the continuous penetration of advanced technologies such as 5G, the Internet of Things and artificial intelligence into the automotive industry, a change in travel mode with automatic driving technology as the core is brewing.
From the first steam driven three wheel vehicle in the world produced by the Frenchman, N. J. Cunnough, in 1769, to the modern automobile with the development trend of automation, electrification, networking and sharing, science and technology has always been the cornerstone to promote the development of the automobile industry. In recent years,With the continuous penetration of advanced technologies such as 5G, the Internet of Things and artificial intelligence into the automotive industry, a change in travel mode with automatic driving technology as the core is brewing.
Automatic driving is an advanced form of automobile intelligence and automation, and a typical representative of the integration of real economy and scientific and technological intelligence. The automatic driving technology will help boost the traditional automobile industry and create new industrial growth points. Countries have successively listed automatic driving as one of the important driving directions.
In China, Baidu, Didi, Tucson Future and other companies are committed to the R&D and testing of autonomous vehicle. In April 2019, China released the Management Specification for Road Testing of Intelligent Connected Vehicles (Implementation). According to incomplete statistics, as of December 2019, about 250 autonomous driving road test licenses have been issued nationwide.
With the development and testing of autonomous vehicle increasing, the acceptance and penetration rate of autonomous driving technology among consumers are also increasing.IResearch predicts that by 2022, the penetration rate of L1-L5 level global autopilot will reach more than 50%, and this number will be close to 70% by 2030.
In March 2019, the Ministry of Industry and Information Technology publicized the draft for approval of Auto Driving Automation Classification (hereinafter referred to as "the draft for approval"), which is planned to be officially implemented on January 1, 2021. The draft for approval clearly divides automatic driving into 6 levels (L0~L5),In the process of the transformation from automatic driving to high-level driving, the role of human beings has gradually changed from driver to passenger.In the process of the realization of L4 and L5 level advanced automatic driving, humans do not need to perform dynamic driving tasks, and do not need to judge the potential safety risks. The auto drive system has the ability to fully take over all dynamic driving tasks. After L5 level automatic driving is realized, Moving Space may be a more appropriate name for a car.
As we all know, the realization of automatic driving requires the system to perfectly depict and predict the road elevation, lane, curvature, dynamic obstacle information, etc. This characterization and prediction requires the close cooperation of vehicle sensors, millimeter level radar, visual computing and other modules, and high-precision maps play a vital role in almost all of the vehicle perception solutions.
In addition, high-precision maps can not only help automatic driving vehicles to obtain the relative position relationship with the external environment, but also can obtain the current status of the vehicle and the absolute position relationship based on the map level, and then compare the perceived position relationship with the real-time data recognized by the sensor to achieve automatic driving.
High precision map - the backbone standing at the center of change
High precision maps include detailed road model, navigation, environmental awareness and other layer information, serving the decision-making and planning layer, and are an important basis for high-level automatic driving.
There are mainly two mapping schemes for refined maps: centralized mapping and crowdsourcing mapping.The centralized mapping mainly depends on professional acquisition vehicles, equipped with high-precision laser radar, GNSS (Global Navigation Satellite System) and wheel speedometer, and its mapping accuracy can be less than 10 cm; With the positioning function of map GPS, the accuracy can even reach 5cm. However, high-precision instruments require higher mapping costs, so they are more suitable for large manufacturers such as Baidu, Gaode, and Siwei Tuxin. Startups have more chosen crowdsourcing mapping schemes with lower costs, and adopted visual schemes, such as cameras and cameras, to replace professional collection vehicles. Although the accuracy of static data cannot be compared with centralized collection, it can still be used as a supplement to dynamic data.
In practical application, high-precision maps also need to be effectively integrated with sensor data carried on autonomous vehicle, and rely on the massive data collected by sensors to compare with high-precision maps, so as to complete a series of actions such as obstacle avoidance, lane change and braking, and improve the accuracy and reliability of automatic driving decisions.
However, the sensor does not have the function of over the horizon, nor can it see the scene behind the object. In rainy or snowy weather, the working state is not so unsatisfactory. Therefore, high-precision maps also provide powerful complementary functions for sensors. Through the data source of high-precision maps, advance environmental awareness is achieved, providing necessary redundancy for sensors. That is, the real-time data update of high-precision maps can ensure that the vehicle can temporarily replace the sensor even if the sensor suddenly fails and its position and environment cannot be confirmed.Therefore, high-precision map+sensor+real-time update may be a solution that can give more energy to automatic driving.
High qualification threshold for high-precision map industryAs of 2019, only a few enterprises in China, such as 4D Tuxin and Changdi Wanfang, are engaged in high-precision mapping.The overall industry presents a tripartite situation of four-dimensional innovation, Changdi Wanfang and Gaode, with the underlying investors corresponding to each othertencent, Baidu andAlibaba, which also reflects thatinternetThe giant is very interested in and optimistic about the high-precision map market.
Looking at the global market, HERE Technologies is a world famous open location service provider. Since 1985, it has been committed to the production of digital maps. In 1984, HERE collected the first electronic map in BMW, making the world's first network map, mobile phone map and high-definition map. In 2016, HERE and Sivetuxin established a joint venture to bring location solutions to China with the help of the powerful power of Chinese enterprises. On July 1, 2020, HERE said on its official website that the new active driving assistant function Active Drive Assist of Ford Motor can rely on the location and judgment of HERE Workspace to achieve automatic driving on highways with a total of more than 100000 miles in the United States and Canada in areas where automatic driving is possible.
At present, HERE can collect 700000 data points every second and 28TB data every day through radar collection. Its unique advantage is that it can directly take the sensor data from the car factory and turn it into products and solutions for the car factory. The car factory can use HERE Platform to let the car factory and users import their own data and create new layers and services, Realize the sharing of sensor data among manufacturers.
In terms of scenes, HERE focuses on high-precision positioning, ADAS networking, electronic horizon and automatic driving of expressways in the short and medium term. HERE adopts a positioning scheme with high-precision GNSS+high-precision maps. It has a very mature dynamic service perception such as ADAS, lane attributes and traffic signals. It can support L2 attributes, L3 and even higher automatic driving levels.
At present, domestic and foreign map dealers are still a long way from mass production of high-precision maps. How to expand the data dimension in the existing situation, how to enable automatic driving to achieve and support more scenarios in a larger range, and what kind of consensus needs to be reached are topics discussed by the entire high-precision map and automatic driving industry.Whether it is to improve the safety of automatic driving from 98% to 100% to achieve qualitative change, or to achieve "twice the driving safety of people", it is the only way for the future development of the global high-precision map and automatic driving industry.