ADAS and Autonomous Driving Industry Chain Report 2018 (VII) - L4 Autonomous Driving Startups at 205 pages in length focuses on researching L4 autonomous driving startups as well as HD map and V2X for L4 autonomous driving.
Of the report series (seven reports), the previous five introduce commercialized ADAS, vision, automotive radar, computing platform, system integration, and low-speed autonomous driving which is to be commercially available soon. The last two reports highlight eventually to-be-commercialized commercial vehicle automated driving and L4 passenger car autonomous driving, respectively.
There have long been two camps in the implementation path of automated driving: Camp A mainly comprised of European and Asian OEMs advocates a progressive path evolving from L2 and L3 to L4 and L5 step by step; Camp B represented by Google stands for a radical path going straight to L4 and above.
In 2018, Camp A believes more firmly that L3 cannot be avoided and L2.5 and L2.75 should be derived from between L2 and L3, and L3.5 from between L3 and L4. To secure the reliability of human and computer driving together, it becomes an important subject to monitor human driver.
Camp B is more confident as well, as WAYMO sees its market capitalization climb to USD175 billion and tests tens of thousands of self-driving cars on roads.
The operational design domain (ODD) of WAYMO self-driving car is confined to just hundreds of square kilometers for the moment; L2-L3 self-driving cars at Camp A can travel on most roads. So the two camps will continue to live in peace with each other in the short run.
In July 2018, John Krafcik, WAYMO’s CEO, admitted that it would take a longer time than expected for the prevalence of autonomous vehicles.
There are at least four technical barriers needing to be surmounted in pushing ahead with L4 from designated scenarios to public roads: first, mass-production of powerful computing platforms; second, stronger sensing capabilities and lower cost of sensors; third, improvement of related technical standards; fourth, inadequate infrastructure. L4 automated driving start-ups will still depend on raised funds to survive in the next two to three years.
We have discussed computing platform and sensor in the previous reports. But L4 development will affect the existing landscape of sensor companies.
Considering too high sensor cost, WAYMO develops by itself all sensor systems it needs, including LiDAR. GM Crusie bought Strobe, a LiDAR company, and Ford Argo acquired Princeton Lightwave, a company engaged in LiDAR. WAYMO can cut 90% cost by developing LiDAR independently; GM Cruise indicates that it can use Strobe’s system to integrate all sensors into one chip, lowering LiDAR cost by 99%.
In addition to sensors, the automated driving leaders also design core computing chips themselves, for example, WAYMO, Tesla and Baidu are all developing their own AI-powered chips.
Singulato, an emerging Chinese automaker indicates that: conventional automotive design is a kind of separate design when it comes to intelligent driving capabilities, that is, separate data cannot be combined for multi-scenario application. In other words, a front ADAS company has a set of sensors of its own and another automated parking company also uses different sensors from others. They cannot share sensor data, which means the waste of resources. Singulato adopts integrated design at the beginning, using same sensors to implement more than a dozen of ADAS functions. And such design also makes subsequent OTA update easier.
Against the backdrop of growing integration, traditional ADAS and sensor companies need to rethink their market orientation in an era of L4.
The number of sensors grows to a dozen and even dozens in the evolution from L2 to L4, generating a data traffic surge. Improvement in supporting facilities, mainly a better perception system, includes introduction of HD map and V2X, which also bring about massive data flow. Data confluence of various perception systems make acquisition, fusion and processing of autonomous driving data flow a focus in industrial competition and cooperation.
Absence of a universally accepted standard for acquisition and transmission of sensor (including HD map) data hinders the development of the industry. Hence, standards organizations like ADASIS, SENORIS, SIP-ADUS, CAICV HD MAP WG and ONEMAP have been initiated.
The year 2018 sees continued improvement in autonomous driving industry chain and influx of capital. As the market prospects of L4 become more visible, HD map and V2X, the auxiliaries of L4, are chased by enterprises and capital.
The author tries to make an overall view of several hundreds of enterprises in autonomous driving industry and present a full picture of the industry via seven industrial-chain reports, 1,400 pages in total, whilst many problems are found, such as irrational layout, unclear orientation, disconnection from industrial chain, and lack of security policy.
As shown in the following diagram, the autonomous driving industry chain is so complicated that it’s a challenge for any enterprise to have a overall grasp of development trends.
Dozens of times larger than the L2 market, the L4 market will take more than five years to grow mature in China. Tracking autonomous and ICV industry, the author will release a weekly report every week and ten monthly reports every month, helping enterprises to see where the industry goes, take in competitive landscape, and seize opportunities in intelligent & connected and autonomous driving markets.
*The Chinese Version of this Report is Available on Request.
1 Overview of L4 Autonomous Driving
2.1.1 Development Course
2.1.2 Investments in Autonomous Driving Field
2.1.3 Large-scale Testing and Verification of Autonomous Driving Safety
2.1.4 Autonomous Driving Simulation System Carcraft
2.1.5 Autonomous Driving System Composition
2.1.6 Computing Platform
2.1.8 Operation of Driverless Taxicab
2.2 GM Cruise
2.2.1 Cruise Autonomous Driving System
2.2.2 Basic Modules of Cruise AV
2.2.3 Deployment of Major Sensors on Cruise AV
2.2.4 Testing Projects and Production Bases
2.2.5 Layout in Autonomous Ride-sharing Mobility
2.3.1 Configurations of Autonomous Vehicle
2.3.2 Three Stages for the Landing of Autonomous Vehicle
2.4.1 RoboCar MiniVan
2.4.2 RoboCar MV 2
2.4.3 Revenue in 2017-2018
2.4.4 Entry to Chinese Market
2.6.1 Acquisition of Princeton Lightwave
2.7.1 Key Technologies and Products
2.7.2 Product Strategy for Autonomous Vehicle
2.7.3 HD Map
2.8.1 Development History
2.8.2 Key Technologies and Tests
2.8.3 R&D and Operation Layout
2.9.1 Core Team
2.9.2 Key Technologies
2.9.3 Development Course
2.9.4 Future Planning
2.10.1 Founding Team
2.10.2 Trend of Development
2.11.1 Key Technologies
2.11.2 Major Products
2.11.3 Development Strategy
2.12.1 Global Presence and Partners
2.12.2 Financing and Development History
2.12.4 Technical Features
2.12.5 Timeline for Autonomous Driving Capabilities
2.13.1 Development Course
2.13.2 System Architecture and Business Model
3.1.1 ADAS Map
3.1.2 HAD Map
3.1.3 HD Map
3.1.4 Dynamic Map
3.1.5 Diversified Forms of HD Map
3.2 Role of HD Map
3.2.1 HD Map Applied in Vehicle Positioning
3.2.2 HD Map Applied in Path Planning and Perception
3.2.3 Role of Dynamic Map
3.3 Standards about HD Map
3.3.1 Autonomous Driving Data Chain and Ecology
3.3.2 Constitution of Autonomous Driving Data Chain Standards
3.3.5 ADASIS V3
3.3.7 CAICV HD MAP WG
3.4 Production of HD Map
3.4.1 Production Flow of HD Map
3.4.2 Data Production of Static Map
3.4.3 Data Updates of Dynamic Map
4.1.1 Development Course
4.1.2 Layout in Automotive Field
4.1.3 Here HD Live Map
4.1.4 Here OTA Solutions
4.1.5 Self-learning HD map
4.1.6 Here and OneMap
4.1.7 Expansion in HD Map
4.2.1 Global Footprint
4.2.3 Revenue Structure
4.2.4 Automotive Business
4.2.5 Telematics Business
4.2.6 Map Development
4.2.7 HD Map
4.2.8 Layout in HD Map and Expansion
4.2.9 To Solve the Problem of Occupant’s Carsickness in Autonomous Vehicle
4.3 AutoNavi (amap.com)
4.3.1 Hierarchical Acquisition System of HD Map
4.3.2 HD Map Data Acquisition Car
4.3.3 HD Map Technology Roadmap
4.4 Baidu Map
4.4.1 HD Map Business
4.4.2 Apollo HD Map File Structure
4.4.3 Apollo Real-time Relative Map
4.5.1 Development Course
4.5.2 Global Customers
4.5.3 Telematics Business
4.5.4 Development Path of HD Map Business
4.5.5 Status Quo of HD Map
4.5.6 Technical Solutions to HD Map
4.5.7 Data Specifications for HD Map
4.6 KuanDeng Technology
4.6.1 Technical Solutions
4.6.2 HD Map
4.7 Deep Map
4.7.1 Financing and Products
4.7.2 Technical Solutions to 3D Map
4.8 Civil Maps
4.8.1 Technical Solutions to 3D Map
4.8.2 Cooperation with Arm in Autonomous Driving Navigation and Positioning Solutions
4.8.3 Cooperation with Renovo
4.9 lvl 5
4.9.1 HD Map Drawing Scheme
4.9.2 Three Levels of lvl 5 HD Map
4.10.1 Partners and Cooperative Projects
4.10.2 Autonomous Vehicle 3D Map Solutions
4.14 Wuhan KOTEI Big Data Corporation
4.15 Qianxun SI
4.16 Dynamic Map Planning
5.1.1 Why to Need V2X
5.1.2 Key Technologies for Vehicle Communications
5.1.3 The Architecture of V2X Communications
5.1.4 V2X Ecosystem and Standards
5.1.5 Countries’ Support for V2X Industry
5.1.6 V2X Use Cases
5.1.7 C-V2X Experiments Worldwide
5.1.8 China’s First V2X Application Layer Group Standard
5.2 Development Stages of V2X
5.2.1 Timeline for V2X Applied in Autonomous Driving
5.2.2 Progress of 3GPP V2X Standards
5.2.3 The First Stage of 3GPP V2X
5.2.4 The Second and Third Stages of 3GPP V2X
5.2.5 Timeline for C-V2X (V2V/V2I) Deployment
6.3 Cohda Wireless
6.5 NEBULA Link
- AutoNavi (amap.com)
- Baidu Map
- Civil Maps
- Cohda Wireless
- Deep Map
- Dynamic Map Planning
- GM Cruise
- KuanDeng Technology
- NEBULA Link
- Qianxun SI
- Sierra Wireless
- Wuhan KOTEI Big Data Corporation
- lvl 5