On May 22,2024, the 2024 sixth Intelligent driving Map and Positioning Conference, jointly sponsored by Galaxy Automobile and the State key Laboratory of surveying and Mapping and remote Sensing Information Engineering of Wuhan University, was successfully concluded in Shanghai, online and offline simultaneously.
With the rapid development of self-driving technology, high-precision map and positioning technology has become the key driving force of intelligent vehicles.
With high precision, high richness and real-time performance, high-precision map provides vital road information for L3 and higher-level autopilot systems.
However, the challenges of cost, coverage and data freshness limit its large-scale application in the field of urban NOA.
Against this background, the sixth Autopilot Map and Positioning Conference 2024 will focus on the key technologies, development directions and latest industry standards of map and positioning, from the subdivided fields of high-precision map data processing, GNSS modularization technology, high-precision IMU, P-Box integrated positioning system, and SD pro maps, and invite a number of guests from different fields of industry, learning and research to talk about the boosting effect of map and positioning technology on autopilot landing.
Promote the innovation and development of high-precision map and positioning technology.
Special thanks to 19 ecological partners, including Beiyun Technology, MCT Smart, Bank of America Technology, OxTS (Oxford Technology Solutions Co., Ltd.), Daoyuan Electronics, AAC Technology, Quick Intelligence, Sinan Navigation / Qintian Navigation, International Navigation Technology, Beijing Rilno, Shitu Technology and so on.
A brief Analysis of China’s passenger car Market & Welcome speech, the domestic passenger car market follows the three stages of technology introduction, independent innovation and development and rapid rise, and maintains rapid growth from 2001 to 2010.
Since 2018, the overall domestic market has shown structural adjustment and steady growth.
Zhou Xiaoying, CEO of Global Automotive, said that at this stage, the transformation of car companies to electrification will determine their future on the Chinese track, while the level of intelligence will determine whether consumers’ experience is good or not.
In mature markets, better quality becomes more important.
The automobile industry is a safety-oriented industry, which should provide safer and reliable products rather than fast-consuming cars to ensure that consumers get a high-quality experience.
At present, the market continues to present new phenomena of diversification, including the iteration of new technologies, the innovation of business models and the deep integration of automobile enterprises and ecological chain partners.
At the same time, the marketing means are becoming more and more abundant.
Auto company leaders actively embrace live broadcast and IP, user-centered, reshape the sales model.
From the traditional dealer channel to APP and direct selling, customer problems can be fed back directly to the core of the enterprise, which greatly accelerates the speed of response and product iteration.
In order to adapt to this change, the supply chain and related dimensions must work closely together to meet the new challenges of the market.
, Zhou Xiaoying | Galaxy CEO, perception, map and location: the three pillars and challenges of building intelligent driving.
As a system platform and the brain of self-driving vehicles, high-precision maps integrate vehicle networking information and perceptual information, directly participate in vehicle control and output navigation decisions.
Its role in intelligent driving can be divided into different levels.
Ying Shen, dean and professor of the School of Resources and Environmental Sciences of Wuhan University, mentioned that the high-precision dynamic map model consists of four layers: static data layer (road network system), road real-time information layer (congestion, events), vehicle self-test and surrounding state layer, and user model layer.
This model integrates static, dynamic and user information, fully covers people, cars and roads, and pays special attention to travel requirements such as comfort, which not only provides the basis for the existing road finders, but also supports intelligent driving decisions.
In addition, the three pillars of smart driving are maps, location and signals.
Among them, the car-road cloud system provides certainty through signals to make up for the perceived uncertainty.
However, smart driving still faces random and uncontrollable “variables” such as weather change, which involve technical, ethical, moral, safety and traffic order.
Therefore, the development road of smart driving is still full of challenges and requires continuous efforts and innovation.
, Ying Shen | Dean, Professor, School of Resources and Environmental Sciences, Wuhan University, the development trend of autopilot high-precision maps.
From January to March 2024, the number of L2 and above intelligent driving machines has reached 1.
94 million, and the penetration rate has reached 43%.
In terms of function, the penetration rate of pilot assist function (NOA) has increased significantly, which has exceeded 5%.
In this regard, Wang Xianbin, senior director of the Global Automotive Research Institute, said that among the more than 200000 models, the standard rate of L2 autopilot has reached nearly 90%, and the future growth potential is mainly concentrated in models below 200000.
Judging from the current auto show and the technical solutions of suppliers, the introduction of pure vision and cost-effective sensor solutions will accelerate the introduction of less than 200000 vehicles into auxiliary driving functions.
Wang Xianbin also pointed out that light maps do not mean that maps are not needed at all, but that the accuracy and dependence of maps are optimized.
For smart cars, navigation maps are still necessary.
The idea of light map is to use other sensors or inertial navigation systems to weaken the dependence on high-precision maps and enhance the perception of specific scenes (such as traffic lights) to meet the high-level needs of intelligent driving.
Wang Xianbin | Senior Director of the Global Automotive Research Institute, map helps smart driving into the second half.
At present, low-level intelligent driving is close to popularity, and the intelligent driving competition is gradually entering the second half-the competition of high-level intelligent driving.
At this stage, the map, as the core support of high-order intelligent driving, has become an important driving force for intelligent driving to enter the second half through continuous iterative updates.
In 2023, with the large-scale application of urban NOA function, a number of OEM and L2mom + intelligent driving domain control manufacturers put forward the slogan “heavy perception light map”, forming a new market wind direction.
Zhang Yang, chief of the autopilot perception section of the Beijing Automobile Research Institute, said that static maps are the main types of high-precision maps in mass production at present.
High-definition maps will overlay dynamic layers on static layers in the future.
The ultimate form of high-precision map-“vehicle collaboration”, that is, the use of advanced wireless communication and new generation Internet technology to implement the dynamic real-time information exchange between vehicles and roads in all directions, and to carry out vehicle active safety control and road collaborative management on the basis of full spatio-temporal dynamic traffic information collection and fusion, so as to fully realize the effective coordination of people, vehicles and roads and ensure traffic safety.
In order to form a safe, efficient and environmentally friendly road traffic system.
Zhang Yang | Chief of automatic driving perception Section of Beijing Automobile Research Institute, the technical trend of high-precision positioning of cars.
Beiyun Technology founder and CEO Xiang said that high-precision positioning in the automotive field is showing three significant trends: first, the development of P-Box to P-MOUSE.
To realize the integration of high-precision positioning and antenna.
second, the globalization of intelligent driving promotes the development of positioning technology to L-BAND or PPP services, pursuing the overall solution of a single chip.
and third, with the development of L3 autopilot, functional safety has become an important consideration of high-precision positioning to meet the functional safety requirements of L3.
The application of automobile high-precision positioning technology is not only limited to intelligent driving, but also suitable for cockpit and other fields.
In the cockpit, high-precision positioning combined with LBS services and lane-level maps can significantly improve the user experience.
With the reduction of the cost of high-precision positioning and high-precision map, this technology is expected to become a standard configuration of cars in the future, further improving the driving and navigation experience of users.
, Xiang is | founder and CEO of Beiyun Technology, Zhongyuan accelerates the landing of full-scene intelligent driving, and there are still many complex scenarios to be solved in the intelligent driving scene, such as invalid location of complex intersections, invalid lane selection in front of intersections, invalid direction selection of similar intersections, and so on, which will be further expanded with the increase of user frequency.
Kang Jian, director of technology marketing and ecology at BofA, said that in order to meet the above challenges, BofA combined with a large number of smart car map application practices to develop production-level spatio-temporal data source systems to provide impetus for the development of intelligent driving.
The system has the characteristics of over-the-horizon, a priori, self-expansion, high freshness, integration, evolution, cost reduction and efficiency, certainty rights trading and other characteristics, not only the introduction of all kinds of scene intelligent driving function, more importantly, continuous experience improvement.
The real large-scale use of users and the establishment of closed-loop data can jointly drive the improvement of intelligent driving experience.
Kangjian also pointed out that no matter with or without maps, heavy maps or light maps, global maps or road features, for vehicle location or cloud training, there is no one or the other, and application is the best.
, Kangjian | Director of Technology Marketing and Ecology of Bank of America, data-driven, software-hardware combination, ushering in a new era of absolute positioning and attitude awareness.
MCT is an innovative company focused on intelligent space-time sensing chips and solutions.
MCT CEO intelligence can mention that in the field of intelligent driving, the application of large language models in the cabin domain, the use of end-to-end perceptual networks at the visual level, the rise of generative artificial intelligence such as ChatGPT, and the landing of closed-loop simulation technology are significant trends in the industry.
What is particularly important is that the way of data acquisition is changing from traditional collection to generation, and this data-driven trend has been very obvious in other sensor fields.
The evolution of high-precision maps has also attracted much attention, from high-precision to light, then to no map, to end-to-end, the map scheme continues to evolve, among which, positioning technology and attitude perception are very important to improve the reliability of L2 + intelligent driving vehicles in China.
At present, the era of location awareness is evolving from chip + offline algorithm to data + AI closed loop.
, MCT / CEO, large-scale application technology of intelligent driving positioning system, Dr.
Xing Li, chief engineer of positioning algorithm of SAIC, believes that the development trend of intelligent driving positioning technology can be summarized in nine words: relying on satellite, focusing on vision and neglecting inertial navigation.
In addition to the well-known RTK services, satellite positioning will also use the relatively low-cost RTD, as well as satellite-based enhanced technology, global coverage of PPP-RTK services.
In the future, the positioning of intelligent driving will pay more attention to vision, using BEV model to build a real-time local map, matching with the base map to achieve accurate positioning, and vision can also be multi-functional.
In the aspect of inertial navigation, the degree of dependence will be reduced and assisted by other sensors.
With the continuous promotion of the scheme of “heavy perception, light map”, the establishment of custom layer based on the underlying map data has become a mainstream development trend. Dr. Xing Li | head engineer of positioning algorithm of SAIC Group, Lang GE Zhijia Map and Zhou Tian Intelligent system, Lang GE Technology Co., Ltd. was founded in 2021, committed to the establishment and operation of map data acquisition, production system.
Relying on tens of millions of Geely production vehicles and millions of Geely travel platforms to collect data around the clock, through deep learning, image recognition, 3D vision, intelligent robot map construction technology and big data cloud based on this, to create a multi-data acquisition, data processing, data production, data service full-link intelligent map data platform.
According to Yang Chuan, head of the Lang GE Science and Technology Intelligent driving algorithm Department, the Lang GE smart driving map data model has four unique coatings, among which the basic layer and update layer rely on Geely car production data to build a high-precision foundation and efficient update mechanism to empower intelligent driving.
OEM operation layer and smart driving experience coating, through the mass production car feedback trajectory and other data to create empirical speed and guide lines, and cover dynamic ODD data to improve the safety and availability of smart driving.
Yangchuan mainly focuses on Lang GE Zhizhi map products, Sunday data intelligence system, intelligent driving truth marking service, and lane-level navigation system.
Yang Chuan | head of Lang GE Technology Intelligent driving algorithm Department, strong enabling of low-cost maps to accelerate the landing of practical application scenarios of self-driving.
Yang du, a polar krypton car navigation expert, said that in the initial stage of self-driving, real-time perception is weak, and high-precision maps have become a “crutch” to achieve self-driving, and complete map elements assistant NOA development.
But the problems of cost and freshness cannot meet the broader demand for autopilot.
With the popularity of algorithms such as BEV+Transformer and occupation network, the real-time perception ability of the vehicle end is enhanced, and the dependence on the map is reduced, and the vehicle only needs key information such as lane topology, slope curvature and so on.
In the future, the ability of real-time perception will continue to improve, and the dependence on maps will continue to decline.
Only SD navigation maps are required to provide a priori information such as global navigation planning and traffic rules, forbidden rules and so on.
The reduction of hardware cost, the weakening of map dependence and the increase of algorithm complexity have become a new trend of map location.
Location confidence will be combined with environmental perception depth to improve the accuracy and reliability of location.
Intelligent driving still depends on the road elements and topology information provided by high-precision maps, but it is more likely to use local location and perception.
Yang du | Polar Krypton car navigation expert, the combination of navigation map and man-machine driving scene, Liu Kunlun, head of the Lailai automobile navigation application team & expert and application software developer, believes that navigation data play an important role in guiding self-driving.
The navigation data of Xilai car not only contains route planning information to guide self-driving vehicles to follow the preset path, but also contains the data needed for lane change decision in advance.
In addition, the EHB data service of Xilai provides the front for self-driving2-3 Information such as kilometer road geometry, road shape points and map line speeds helps them achieve early lane change and pre-deceleration.
Lane line information helps the autonomous driving system decide on lane change timing and available lanes.
Real-time traffic information updates congestion and road space-time information in real time to further enhance the intelligent driving experience.
, In terms of positioning matching, NIO Automobile checks the real-time positioning results of the map with the positioning services of the autonomous driving system to ensure the accuracy of positioning.
In addition, NIO Automobile uses Ethernet to realize data interaction between the smart driving system and the side system at a rate of hundreds of Hertz per second, with a daily data volume of about 5 megabytes to ensure information flow between systems.
Liu Kunlun| NIO’s car navigation application team leader and expert, application software research and development, and the application of small high-performance, low-power IMU and GNSS/INS products in the fields of surveying and mapping and high-precision positioning.
Honeywell has been involved in navigation and autonomous driving since 1914.
In the field of driving, with more than a hundred years of technical accumulation, it has become a pioneer in aerospace inertial navigation technology.
Honeywell has always maintained technological iteration and innovation and continues to lead the field of autonomous driving.
, Zhou Jianyu, Director of Commercial and General Aviation Business for Honeywell Aerospace Technology Group Asia Pacific, gave a detailed introduction to Honeywell’s navigation and autonomous driving products.
HGuide O360 is an onboard INS/GNSS navigation product that meets the needs of high-end autonomous driving.
Integrate data from onboard IMU and GNSS receivers with other sensors (DMI, DVL, etc.
) to provide reliable and robust navigation services.
The O360 can be easily docked with lidar and is specially designed to accommodate a wide range of vehicle platforms.
, Zhou Jianyu| Director of Business and General Aviation Business for the Asia-Pacific Region of Honeywell Aerospace Technology Group.
At this point, the entire agenda of the 2024 6th Autopilot Mapping and Positioning Conference has been successfully concluded.
As one of the core supports of autonomous driving technology, high-precision maps are directly related to the safety and reliability of autonomous driving vehicles.
In the future, we expect high-precision map technology to continue to innovate and deeply integrate with autonomous driving technology to jointly promote the automotive industry in the direction of intelligence, safety and efficiency!, Return to First Electric Network Home>.
