RCCN WeChat QrCode
Mobile WebSiteAutomatic driving is the automobile industry and artificial intelligence, Internet of things, high-performance computing and other new generation of information technology deep integration of the product, is the current global automotive and transportation travel intelligent and network development of the main direction, has become a national competition Strategic high ground. This article focuses on the technologies and standards involved in automatic driving, as well as the progress and trends in testing and deployment at home and abroad.
Analysis on the Current Situation and Prospect of Automatic Driving Technology
First, automatic driving technology and classification
(A) automatic driving technology classification
Automatic driving technology is divided into multiple grades, the current domestic and foreign industries with more for the American Society of Automotive Engineers (SAE) and the US Highway Safety Administration (NHTSA) introduced the classification criteria. (L0), driving support (L1), partial automation (L2), conditional automation (L3), highly automated (L4), automatic automation (L4), automatic automation (L4) ) And fully automated (L5). The main difference between the two different classification criteria is that under full autopilot, SAE subdivides the scope of the autopilot system. Detailed standard see below:
(B) automatic driving technology route
In the automatic driving technology, there are two different development routes:
The first is the "gradual evolution" of the line, that is, in today's car gradually add some automatic driving function, such as Tesla, BMW, Audi, Ford and other car prices are used in this way, this way the main use Sensor, through the car communication (V2V), car cloud communication to achieve road analysis.
The second is a completely "revolutionary" line, that is, from the beginning is a thorough autopilot car, such as Google and Ford is a number of structured environment in the test of the automatic driving car, this route mainly rely on car laser Radar, computer and control system to achieve automatic driving.
From the application scenario, the first approach is more suitable for testing on structured roads, the second way in addition to structural roads, but also for military or special areas.
(C) automatic driving involved in the hardware and software
1, the sensor
The sensor is equivalent to driving the car's eyes. Through the sensor, the autopilot can identify roads, other vehicles, pedestrian obstructions and basic transport facilities. According to the automatic driving of different technical routes, the sensor can be divided into three types of laser radar, traditional radar and camera.
(1) Lidar
Is currently the largest proportion of equipment, Google, Baidu, Uber and other companies of the company's automatic driving technology is currently dependent on it, this device was racked on the roof of the car, the laser pulse to the surrounding environment for distance detection, and Combined with software to draw 3D images, so as to provide enough automatic driving vehicle environmental information. Lidars have the ability to accurately and quickly identify, the only drawback is the high cost (average price of 80,000 US dollars a) lead to mass production cars difficult to use the technology.
(2) traditional radar and camera
As a result of the high price of laser radar, take the practical technology of the car prices have turned to the traditional radar and camera as a sensor alternative, such as the famous electric car manufacturer Tesla, the program is the radar and monocular camera, internationally renowned manufacturers Mobileye et al. The hardware principle is similar to the current ACC-compliant cruise system of the vehicle, relying on a camera and a front radar covering the 360 ° viewing angle around the car to identify the three-dimensional information, thus ensuring that the vehicles do not collide with each other.
Although this sensor program is low cost, easy to mass production, but for the recognition of the camera has a high demand: monocular camera need to establish and maintain a large sample feature database, if the need to identify the target feature data, The system can not identify and ranging, it is easy to lead to accidents. The binocular camera can be directly on the front of the scene ranging, but the difficulty lies in the calculation of large, need to improve the performance of computing units.
2, high-precision map
Automatic driving technology for lanes, car distance, roadblocks and other information more dependent on the need for more accurate location information, is the basis for automatic driving vehicle to understand the environment, with the continuous evolution of automatic driving technology upgrade, in order to achieve the safety of decision-making , Need to reach the centimeter-level accuracy. If the sensor provides an intuitive environmental impression to the autopilot vehicle, the high-precision map can be accurately positioned by the vehicle to accurately restore the vehicle in a dynamically changing three-dimensional traffic environment.
3, V2X
V2X, refers to the vehicle and the surrounding mobile traffic control system to achieve the interaction of technology, X can be a vehicle, traffic lights can be traffic lights, can also be a cloud database, the ultimate goal is to help automate the vehicle to master real-time driving information and Traffic information, combined with vehicle engineering algorithm to make decisions, is the key to driving the vehicle into the unmanned phase.
4, AI algorithm
Algorithm is to support the most critical part of automatic driving technology, the current mainstream automatic driving companies have adopted the machine learning and artificial intelligence algorithm to achieve. Massive data is the basis for machine learning and artificial intelligence algorithms, data obtained from previously mentioned sensors, V2X facilities and high-precision map information, as well as collected driving behavior, driving experience, driving rules, cases and surroundings Data information, and constantly optimized algorithms to identify and ultimately plan the route, manipulate driving.
Second, domestic and international development and trends
From the automatic development of the whole domestic and international development situation, the virtues lead the development of automatic driving industry tide, Japan, South Korea quickly awakened, China was catching up. Specifically, the following trends are reflected:
(A) as soon as possible as the goal of commercial, to speed up the introduction of road tests and regulations introduced
In the road test, the United States, Germany, Japan, Korea, China are actively promoting road test, as the basis for automatic driving car applications. From an international perspective, countries have to 2020 as an important time node, hoping to achieve full deployment of automatic driving vehicles.
At the end of 2016, 16 states passed the relevant laws or administrative orders to clarify test conditions and requirements, allowing companies to conduct road tests at the state level.
The German government has allowed autopilot car testing on the A9 motorway connecting Munich and Berlin in 2015. The transportation sector also subsidized the Diginet-PS pilot project in Berlin in March to develop a processing system and provide Automatic driving of real-time traffic information.
Japan's Nissan has tested its own driving vehicle LEAF in Tokyo, Silicon Valley and London, hoping to accumulate safety test records as soon as possible.
South Korea has issued 13 autopilot test licenses, plans to commercialized by 2020, three automatic driving car.
From the perspective of China, the Ministry of Industry and the Ministry of Industry and Technology in Shanghai in 2010 to carry out the Shanghai intelligent network of pilot demonstration; in Zhejiang, Beijing - Hebei, Chongqing, Jilin, Hubei and other places to carry out "broadband mobile Internet smart car, intelligent traffic application demonstration" Driving test work.
Beijing has introduced a five-year action plan for intelligent vehicle and intelligent traffic application demonstration. It will complete the transformation of all the main roads and roads in the Beijing Development Zone by the end of 2020, and plan the application of 1000 fully automated driving vehicles in phases. Jiangsu in November 2016 with the Ministry of Industry, Ministry of Public Security signed a tripartite cooperation agreement to build a national intelligent traffic comprehensive test base.
(B) to the network as the direction of the car, to promote system research and development and communication standards
From the current industry trends, most companies have adopted the development of Connected Cars, speed up the chip processing capacity, automatic driving cognitive system research and development, to promote the introduction of unified vehicle communication standards.
R & D, Bosch Group and NVIDIA are working together to develop an artificial intelligence autopilot system, NVIDIA provides depth learning software and hardware, Bosch AI will be based on NVIDIA Drive PX technology and the company's upcoming super chip Xavier, then can provide level 4 Automatic driving technology. IBM announced that its scientists have acquired a patent for a machine learning system that can dynamically change the control of autonomous vehicles between human pilots and vehicle control processors in potentially urgent situations to prevent accidents.
In terms of vehicle communication standards, LTE-V, 5G and other communication technologies become the key to the automatic driving vehicle communication standard, which will provide high-speed and low-delay network support for autopilot.
On the one hand, domestic and international cooperation to promote LTE-V2X 3GPP 4.5G important development direction. Datang, Huawei, China Mobile, China Telecom and other efforts to promote the V2V, V2I standardization work has made positive progress.
On the other hand, LTE-V2X technology is also with the development of automatic driving demand is gradually evolved to 5GV2X. 5G, V2X dedicated communication can be extended to the scope of the sensor outside the working area of the sensor, to achieve safe high-bandwidth business applications and automatic driving to complete the car from the travel tool to the information platform, entertainment platform transformation, help to further enrich the business situation The
Currently, the 5GA (5GAA) and the European Automotive and Telecommunications Alliance (EATA) have signed a memorandum of understanding that will work together to promote the C-V2X industry, using standardized, spectrum and pre-deployment projects based on cellular communications technology. China Mobile and Beiqi, GM, Audi and other cooperation to promote 5G joint innovation, Huawei and BMW, Audi and other cooperation to promote the development of services based on 5G.
In addition, the Ministry of Industry and the organization of the intelligent network of automotive standard system program will be released, the car network standard system is also gradually improved, for the development of intelligent network of vehicles is essential.
(C) to lead the innovation format, the Internet companies to become an important driving force
Internet companies are born with business innovation and development of the gene, are also involved in the automatic driving industry, has become an important driving force in the industry.
In the United States, Google has started unmanned business research and development in 2009, December 2005 to December 2016 in California on the road driving a total of 635,868 miles, not only the California test mileage of the largest enterprises, but also the system to disable the lowest enterprise. Uber, the largest US provider of car service in the United States, has been unmanned road test in Pittsburgh, Tempe, San Francisco and California. Lyft, the second largest Internet service provider, announced the three-phase development plan for autopilot vehicles in September last year. Has also been tested in Pittsburgh. Apple also in April this year, just won the California test license. South Korea has just approved the South Korean Internet company Naver on the road to test the autopilot car, becoming the first 13 licensed autopilot car R & D enterprises, plans to commercialized by 2020, three automatic driving car.
From our point of view, Baidu company in September last year, won the test permit in California, in November in Wuzhen, Zhejiang to carry out the general open road unmanned car test operation. Its president and chief operating officer Lu Qi is on April 19 this year, released the "Apollo" program, plans to master the company's automatic driving technology to the industry open, the open environment perception, path planning, vehicle control, car operating system Function of the code or ability, and provide a complete development of testing tools, the purpose is to further reduce the unmanned research and development of the threshold to promote the rapid popularization of technology.
Tencent in the second half of 2016 to set up automatic driving laboratory, relying on 360-degree look around, high-precision map, point cloud information processing and integration of positioning and other aspects of technology accumulation, focus on automatic driving core technology research and development. Ali, music, etc. have also been with the SAIC and other car prices to develop Internet vehicles.
(4) to corporate mergers and acquisitions as a breakthrough, start-up enterprises and leading enterprises become the subject
Autopilot The rapid development of the main business of mergers and acquisitions to master the key technology of automatic driving enterprises or start-ups.
In July 2016, General Motors acquired Cruise Automation, a Silicon Valley start-up company, for more than $ 1 billion, with the RP-1 expressway autopilot system with high automation driving potential.
In March 2017, Intel acquired the Israeli technology company Mobileye for $ 15.3 billion, which is dedicated to the development of hardware and software systems related to autopilot, which is the main camera supplier for company driving assistance systems such as Tesla and BMW. Image recognition aspects of the patent.
The company acquired the deCarta, a start-up company that offers location APIs in 2015, and has acquired employees who are proficient in image and data collection from Microsoft's Bing division.
In April 2017, Baidu announced the acquisition of a wholly-owned acquisition of a machine vision hardware and software solutions, the US technology company xPerception, the company for robot, AR / VR, intelligent guide and other industries to provide customers with three-dimensional inertia camera as the core of the machine Visual hardware and software products, intelligent hardware can be realized in a strange environment on their own positioning, the calculation of space three-dimensional structure and path planning. According to industry analysis, Baidu may be in order to strengthen the field of visual perception of hardware and software capabilities.
In general, the acquisition of leading enterprises or potential start-ups, can quickly accelerate the accumulation of their own driving technology, the formation of competitive advantage.
