Creative Pictures/New Beijing News Reporter Wang Yuanzheng
Original Title: Autopilot: Technological commercialization is still a mystery
With the deep exploration of autopilot technology, autopilot practitioners begin to have a more sober understanding of its commercial landing. The development of this technology may be more expensive and complex than initially expected.
In fact, in many aspects, such as vehicle-road collaboration, algorithm and precision map related to automatic driving technology, domestic and foreign auto-driving enterprises have made in-depth attempts, but the results are not optimistic, and their test data are mostly focused on specific test roads, rather than open roads.
Therefore, an engineer who studies self-driving said in an interview with the Beijing News that the commercialization of self-driving may take a long time. This has nothing to do with technology promotion, but lies in openness, which includes open platform, open road, and open legal norms.
New Beijing News reporter combs the current development status and planning of mainstream autopilot enterprises at home and abroad, in order to observe the pace of its progress and the process of commercialization. To be sure, the realization of commercialization may still need to include the formulation and support of industry-wide and national standards.
Domestic representative enterprises
Key words: Apollo
Beginning in 2015, Baidu took the lead in testing and testing unmanned vehicles, researching and reserving the related technology fields of automatic driving. At present, Baidu Apollo has been upgraded to version 5.0. In addition, Baidu has entered into cooperative relations with FAW Group, Chang’an Group, BMW Group, CEF, Horizon and other relevant enterprises in more than 140 industrial chains. In terms of commercialization, in July 2019, Apollo 5.0 was officially mass-produced to achieve autonomous driving in a limited area. Previously, the L4 self-driving taxi Robotaxi-Red Flag E. Boundary, which was built by Baidu Apollo and FAW Red Flag, has officially appeared in Changsha and started testing. According to Baidu’s plan, full-automatic driving will be achieved on high-speed and urban roads in 2021.
It is worth noting that even Baidu is not all a shining moment. The high investment in self-driving has tested all practitioners, and Baidu was also released in May this year to prepare for the breakdown of the plan. Although Baidu denies this statement, repeated divisions and unsatisfactory financial statements have not completely dispelled doubts from the outside world.
Key words: vehicle-road coordination
In the aspect of automatic driving, Ali chose to reduce the physical predicament and cost barrier of the existing scheme of driverless driving in the way of cooperative intelligence, that is, focusing on vehicle-road coordination.
At present, Ali’s advantages are mainly concentrated in the roadside, and has the advantages of Ali Bird Alliance scenario, ET city brain, intelligent transportation and so on. In June this year, Ali jointly launched a smart high-speed solution, which will cover high-speed operation monitoring, abnormal event detection and disposal, and future road condition prediction. By 2020, the highway network of 10 provinces and cities will be landed.
Ali believes that road-to-vehicle collaboration is the ultimate form and future direction of research and development of automatic driving. The core component of road-to-road collaboration is the perception base station, which not only greatly reduces costs, but also improves the safety of automatic driving. Even in the future, it can provide intelligent road information service for ordinary motor vehicles.
Key words: Autopilot Travel Ecology
Tencent Autopilot Business Center was established in 2016. In November 2018, at the Tencent Global Partnership Conference, Tencent announced its first smart travel strategy, and Tencent Autopilot officially appeared. Tencent’s layout in the field of automatic driving is more inclined to the construction of travel ecology.
Tencent believes that in the direction of industrial Internet, car building is not Tencent’s advantage. While conducting large-scale road surveys, the key to realize large-scale commercialization of automatic driving is to focus on the travel ecology in the industrial chain. Based on its own characteristics, Tencent in-depth analysis of vehicle-road collaboration problems that need to be solved, joint industry chain related enterprises, joint research and construction of infrastructure, platform services, business applications, three aspects of vehicle-road collaborative innovation ecosystem.
Tencent Autopilot, which started in 2016, is not too early. The commercial prospect of 5G technology not only brings hope to the whole industry of automatic driving, but also brings a new breakthrough to Tencent’s construction of ecological vehicle network. It is understood that Tencent’s internal 5G research team has carried out 5G vehicle-road collaboration commercial experiments.
But for all participants in the 5G field, whether it is large-scale commercial or real landing of road-to-vehicle collaboration requires time to verify. For Tencent, which started late, it will be a test for its technological research and development ability and synergistic ability with other industries.
Domestic self-driving start-ups
Key words: potential stocks
Xiao Ma Zhihang is a company specializing in autopilot solutions. In January 2018, Xiaoma Zhihang announced the completion of a $112 million round A financing, which is mainly used for team development, new generation of unmanned vehicle platform development, unmanned fleet operation and other directions.
In March 2017, Xiao Ma Zhihang set up the Beijing R&D Center and completed the first self-driving prototype car in Silicon Valley in April of that year. Then he got the California road test license and completed the first full-automatic driving test. At present, Xiao Ma Zhihang has also obtained the T3 level automatic driving road test license issued by Beijing, becoming the only start-up company to get this level license.
In September 2018, Xiaoma Zhihang officially released the third generation of automatic driving software and hardware integrated system configuration Pony Alpha, opening the road of the production of unmanned motorcade. In addition, Xiaoma Zhihang also tailored the system support platform for the fleet operation.
But at the same time, Xiao Ma Zhihang is also facing the problem of commercial landing caused by capital and other issues. “When we invest, we understand that this is a long cycle,” one investor said in an interview with the media. At the technical level, Xiaoma Zhihang has accumulated a certain amount of technology, but a large amount of capital is still needed to realize the commercial operation scenario.
Wen Yuan knows and acts
Key words: multi-sensor fusion
Wen Yuanzhixing was founded in Silicon Valley in the United States in 2017. In December of the same year, he decided to use Guangzhou as the global headquarters. In June 2019, Guangzhou issued the first batch of 24 automatic driving road test licenses, of which 20 were won by Wen Yuan Zhixing.
Since its inception, Wen Yuanyuan has focused on the strategic collaboration of AI technology with automobile enterprises and travel platforms. Its core technology is a multi-sensor fusion scheme which combines camera, lidar and millimeter wave radar to enable vehicles to sense roadblocks, locations, directions and speeds of movement. This involves hardware combination, image processing and Computer vision recognition algorithm. In order to make a good driving plan for the car, Wen Yuanzhi built a high-precision map that no one can understand. This high-precision map is a 3D model and can achieve centimeter-level accuracy.
At present, Wen Yuanzhixing has achieved nearly 500,000 kilometers of road test mileage for self-driving vehicles in Beijing and 560,000 kilometers in China and the United States.
Key words: AI chip
Founded in 2015, Horizon Technology has been positioned to be more than just a chip company. Horizon started as a platform company, focusing on providing the industry with the best AI computing platform.
In 2017, Horizon launched a series of margin AI chips for intelligent driving. The most important feature is to integrate computing features into the design of computing architecture forward-looking. After one or two years of research and development, AI processors are still able to adapt to the latest mainstream algorithms. At present, the Matrix Autopilot Computing Platform based on the Stroke 2.0 framework has entered a large-scale commercial stage.
Up to now, the shipment of AI chips on the horizon edge reached 100,000 pieces in 2018. Over the next two to three years, Horizon plans to achieve tens of millions of shipments. According to the Horizon plan, it will take 5 to 10 years to be one of the leaders in edge AI computing.
Four Dimensional Mapping
Key words: big location data
Four-Dimensional Map New is a comprehensive autopilot technology provider, with navigation business, advanced assistant driving and autopilot business, chip business, location data service, and automobile networking business as its main business lines.
In the field of automatic driving, ADAS Map has covered national highways, national highways, provincial highways and important urban roads. Sample data validation is being carried out with a number of customers. Application-oriented products based on ADAS map begin to promote vehicle testing and continuous optimization. The new four-dimensional map has the capacity to support high-precision map of L3 and above Freeway Automatic driving.
In terms of location-based big data business, the new four-dimensional map released MineData 3.0, a location-based big data platform, at the end of July this year. Based on the new platform and visualization capabilities, it continued to optimize the industry-specific solutions, and launched a new overall solution for the transportation, traffic police, public security, planning and insurance industries.
In addition to its own technological system, Four-Dimensional Mapping is also building a wider ecosphere of smart cars through external cooperation, investment and financing. In April this year, Six Technologies, a high-precision positioning service provider under Four-Dimensional Map New, introduced strategic investment from China Telecom and China Internet Investment Fund to provide an integrated high-precision positioning service solution for “Internet Cloud”.
Key words: road test “king”
Waymo is the earliest autopilot player in the world. It was first a company specializing in autopilot research under Alphabet, the parent company of Google, and then served in the autopilot program business of Google. It was split into independent company Waymo in 2016. The split means Waymo is seen as a business project that can be sustained for a long time, and commercialization is on the agenda.
In fact, Waymo has been a pioneer in automated road testing and commercialization for 10 years. According to the February news, the total mileage of Waymo’s autopilot test may have exceeded 20 million kilometers.
In July, Waymo announced that it would open data sets for automatic driving and that it would release more in the future. Openness is also seen as a key to the technological advancement of autonomous driving.
However, with the rapid commercialization, Waymo has also experienced how to solve the moral problems in future traffic, how to balance manual operation and automatic driving.
Key words: capital care
According to Navigant Research, an American industry research company, Cruise ranks second only to Waymo, a Google Brothers company, in the autopilot research and development rankings.
Cruise was founded in 2013 and acquired by GM for $580 million in 2016. It is now the core force responsible for the development of L4 level and above in the GM system. On July 5 this year, Softbank Group of Japan provided an investment plan of $2.25 billion to General Motors subsidiary Cruise, which was approved by the National Security Council of the United States. After the investment is completed, Softbank can win a seat on Cruise’s board of directors and get a 20% stake in Cruise.
Cruise is considered the industry leader and is expected to generate $173.15 billion in revenue by 2023. At present, GM is working hard to prepare for the introduction of automatic taxi service by 2020. According to the latest reports, Cruise, GM’s autopilot subsidiary, has postponed plans to launch an autopilot taxi service by the end of this year. Delaying automated taxi service is similar to what most automated taxi providers are currently facing. The development of this technology may be more expensive, complex and time-consuming than originally expected.
Ford Argo AI
Key words: cooperation
The industry pays much more attention to Ford Autopilot than to Ford itself. The alliance between Volkswagen and Ford has also brought Ford Argo AI into people’s minds.
On July 12 this year, rumors of the Volkswagen Ford Autopilot Alliance have settled for nearly half a year. Volkswagen announced a $2.6 billion investment in Ford-backed autopilot Argo AI, including a $1 billion investment and a $1.6 billion autopilot technology subsidiary, AID. Argo AI’s total valuation rose from $4 billion to more than $7 billion.
With financial and technical support from the public, Argo AI said it would speed up commercialization, conduct technical tests with Ford in Miami and Washington, and deploy commercial services. Volkswagen and Ford will be equipped with Argo AI’s self-driving technology system to better support the planning of their respective personnel and cargo transport areas.
In Argo AI’s eyes, in partnership with Volkswagen and Ford, Argo AI’s technology could cover multiple brands and models in North American and European markets in the future.
Wei Shuai, Reporter of the Beijing News