TU-TESL

• 22

[editor's note] as far as perceptual information is concerned, high-precision map can not only provide high-precision static information for cars, such as road network, shape, lane, POI, building, road sign, etc., but also contain dynamic real-time traffic information. Through the integration of these two kinds of information, a virtual driving environment is formed for cars to perceive, recognize and understand the environment, and carry out path planning Avoid congestion and traffic obstacles. For human drivers, the main function of electronic map is navigation, including path planning from a to B, location matching of vehicles and roads, POI retrieval, etc., so when the future car can achieve a certain degree of automatic driving, or even without drivers, can complete automatic driving, what kind of map is needed? The answer is a high-precision map. It is particularly important to accurately evaluate the location of oneself and perceive the surrounding environment. At present, sensors such as camera, millimeter wave radar and laser radar have some defects in environmental perception, especially in extreme weather such as dust, rain, snow and fog, which are prone to misjudgment or even failure. Even if multiple sensors are integrated, they cannot be completely avoided. The high-precision map can still play a role not only in extreme weather conditions, but also in the field of vision, which is not subject to occlusion, distance and visual limitations. It can form a good complement with the sensor in the sensing layer, and provide a more reliable perception for the self driving vehicle. From this point of view, the high-precision map is actually equivalent to a super perception container. On the one hand, it can assist the existing sensors, on the other hand, it can serve as a platform to meet the needs of lane level planning, and finally realize the double enhancement of perception and decision-making. It is worth mentioning that based on the reconstruction of 3D road environment, high-precision maps can help autopilot reduce the dependence on expensive sensors, greatly reduce the system cost and reduce the computational pressure inside the vehicle. In the view of Lu Zheyuan, Mdt InfoTech Ltd of Shanghai, the high-precision map can help the autonomous driving vehicle to have a location perception and more precise path planning, and provide support for the decision-making level, be of great advantage to the development of intelligent transportation. For example, in the field of intelligent parking, it can be used for parking space guidance and reverse search to help users quickly find parking spaces. "The high-precision map can also be used as the supplement and enhancement of the existing sensors of automatic driving, to strengthen the perception ability of the vehicle end in the vehicle road collaborative architecture, and then improve the intelligent network application. As well as power car enterprises, scientific research institutions, etc. to carry out the virtual test of automatic driving. " Lu said. Guo Panshi, vice president of new car road Collaborative Research Institute of four-dimensional map, believes that compared with the traditional map, the map required by automatic driving has higher requirements in terms of accuracy, which does not mean that the absolute accuracy must reach what level, but rather that the accuracy can cover all kinds of scenes required for travel. "For example, when we drive on a highway, what we need is not the absolute geographic coordinate accuracy of vehicles, but the relative relationship between lane lines." In addition, Guo believes that the map used for automatic driving should also have the characteristics of complete elements, fast update, strong coordination, etc. "What is all elements? For example, if we use didi taxi, you need to know whether the boarding point is the east gate, west gate, south gate or North Gate of the community, which should also be shown on the map. As for the update, although the second level update has been implemented, it is not enough for automatic driving, and it needs to be further achieved. Strong collaboration means that map can cooperate with sensing system, computing system and communication system in an all-round way. " However, given the number of different levels of autonomous driving, not every stage requires such sophisticated technology. According to Liu Bin, policy research center of Automobile Technology Information Research Institute of China automobile center, traditional electronic map is still needed as travel reference in L1 and L2 stages, because it is human drivers who control vehicles at this time. However, at L3 level, due to the control right shared by people and the system, the introduction of high-precision map and the combination of sensors will help to reduce the cost of research and development and facilitate subsequent mass production. Yang diange, head of the Automotive Engineering Department of Tsinghua University, also believes that for the L1 and L2 ADAS systems, the sub meter ADAS map can be used, while for L3, in addition to ADAS map, high-precision map may be used, but it is not necessary. In L4 stage, centimeter level high-precision map is necessary, and the same is true for L5, which not only must have but also can be updated in real time. Compared with the traditional electronic map, high-precision map, because of its higher accuracy and wider coverage, can provide more detailed and detailed environmental information for vehicles. It has become one of the core technologies of autopilot. However, because the data provided by autopilot map is too detailed and involves the security of spatial information, there are many limitations in the current laws and policies in data collection, transmission, storage, use and expression, which to a certain extent restricts the development of high-precision map. Specifically, first, this is mainly for vehicle enterprises and solution providers of automatic driving solutions. Due to the limitations of current laws and regulations, they can't collect, use and store these spatial location information without surveying and mapping qualification, and can only cooperate with qualified cartographers. In addition, it also includes crowdsourcing collection, elevation, slope, curvature of roads, height and weight limits of bridges and tunnels. According to the current policy, there are also clear restrictions. And the car companies have a strong demand for these data, which indirectly affects the development of automatic driving. Second, the current international standards are basically based on European and American road design, which is quite different from the domestic scene. Thirdly, according to the current laws and regulations, autopilot map is still a kind of navigation electronic map, which needs to be processed before it is put into public use, and autopilot has high requirements for positioning, which obviously contradicts the requirements of laws and regulations. Fourth, at present, the navigation electronic map implements the license system, which needs to be reviewed by the competent geographic information department before publication and distribution. However, the automatic driving map uses the form of digital string to express relevant information, and has a high update frequency and cycle. The current review mode is difficult to meet these needs. Fifthly, at present, the domestic automatic driving test field mainly focuses on the test of automatic driving technology, and there is no map related test verification, which leads to the lack of evaluation of geographic information security. Sixth, the automatic driving map not only contains high-precision static information of the road, but also may contain dynamic information such as traffic events and road construction in the future. Based on such a feature, the cost of data acquisition and update will be very high in the future. If there is a unified data management platform for automatic driving data collection, diagnosis and evaluation, it can achieve more efficient data sharing. It can be seen that the industrialization of high-precision map is also a long way. Nevertheless, the field has attracted a large number of enterprises in the past few years. Looking at the market, in addition to the traditional map merchants, such as bat and other technology giants, as well as BBA and other traditional vehicle enterprises, they are all entering the field of high-precision map by means of acquisition, investment or cooperation, and even a large number of start-up enterprises have emerged, such as momenta, Kuan stool, Jingzhong, etc. For bat, Baidu has obtained high-precision maps and self positioning mass production orders from Great Wall Motors, and signed commercial fixed-point agreements with,,, automobile, GAC motor, Dacheng automobile and many other brands. Alibaba's Gaode map has successively obtained and two high-precision map commercial orders, while Tencent's four-dimensional innovation has obtained China's mass production orders. And start-up companies are also constantly strengthening cooperation with car companies to speed up their integration into the market and get more development space. Under the competition of many forces, the market pace of high-precision map has been accelerated. According to the prediction of the geyser Automobile Research Institute,