Technology cannot be kept in a cage forever. At this bumpy node of the road ahead of autonomous driving, this epidemic undoubtedly provides an opportunity for autonomous driving enterprises to "combat", so that autonomous driving can test the perfection of technology and the maturity of products more before it is officially commercialized.
Image source: Apollo official website
Of course, at present, whether it is the development of technology or the feasibility of application, autonomous driving is still too far away from mass consumers. It needs to solve many problems including technology, safety, regulations and cost.
As we all know, autonomous driving relies on various sensors to collect the surrounding information while driving. Although the development of sensors has been extremely rapid, there are still visual blind spots. In addition, 5G technology has not been popularized on a large scale, and cloud data delay will also bring security risks. The deeper problem is that even if it is technically safe, autonomous driving can't really drive on the road. At least at present, the cooperation between self-driving cars and roads in China only stays in the demonstration area, and no local government has really opened the road to self-driving cars.
In addition, the existing laws are not suitable for autonomous driving, especially between L3 and L4, and the boundary between people and vehicles is unclear. Who is responsible for the vehicle accident? These all need perfect laws and regulations to explain. The most realistic problems are cost and profit. Software and hardware facilities such as lidar and chips have not been mass-produced, and it is quite expensive for car companies to purchase related equipment ... In addition, the construction of supporting infrastructure including high-precision maps and autonomous driving operations still faces great challenges.
Fortunately, in recent years, with the cooperation of the government, enterprises and other parties, the problem is gradually being overcome. ?
The key technology of mass production has gradually matured.
For autonomous driving, the technology at the sensor and map level is particularly critical.
As the "eyes" of autonomous driving, there are many kinds of sensors, such as camera, ultrasonic radar, lidar and millimeter-wave radar, among which lidar plays a very important role. It is not affected by environmental factors such as light, and can accurately obtain the three-dimensional position information of the front object and identify the surrounding information with high precision. However, due to the high cost of lidar, there is no way to mass-produce.
However, in recent years, with the maturity of lidar technology and the entry of a large number of technology companies, the price of existing lidar has been as low as 65,438+000 dollars. Although the performance is not as high as tens of thousands of dollars, it is not easy to achieve such a price. More importantly, the industry has seen the possibility of reducing the cost of lidar. In the past, the difficulty of mass production and cost reduction of lidar may not be so difficult. In the future, the price of lidar can be expected to drop, reaching the mass production level of $65,438+000-$65,438+0000, which may not be far off.
Image source: Wilton? laser radar
In addition to sensing technology, accurate positioning is not only the basis but also the core of autonomous driving. Therefore, high-precision maps have become a hot "man of the hour" in the field of autonomous driving. In addition to traditional map manufacturers, technology giants such as BAT and traditional car companies such as BBA are entering the field of high-precision maps through acquisition, investment or cooperation, and even a large number of start-ups such as Momenta, Broad Stool and Zhong Jing have been born. At present, more than 20 enterprises have obtained domestic high-precision surveying and mapping qualifications, among which Huawei, SF Express and Jingdong Logistics are the latest entrants. It is believed that as more and more enterprises settle in, the accuracy of high-precision maps will be improved more obviously, and there will be more and more applicable scenarios.
The system of laws and regulations is becoming more and more perfect.
With the maturity of autonomous driving technology, relevant autonomous driving standards are also on the road, and autonomous driving is not without "standards".
Not long ago, South Korea took the lead in issuing safety standards for self-driving vehicles, which mainly included: automatic lane keeping for L3-class self-driving vehicles, monitoring the driver's condition in case of emergency during driving, and auxiliary functions such as automatically slowing down, starting emergency warning signals and reducing danger when there is no timely manual response. Therefore, South Korea became the first country in the world to formulate L3-class self-driving vehicle safety standards and commercialization standards.
Of course, in addition to South Korea, the US Department of Transportation (USDOT) has also released the latest self-driving car guide 4.0(AV? 4.0), which is the fourth version in the United States since 20 16. Other western countries are also improving the relevant legal standards: Spain 20 14? The traffic regulations were revised in 1995; In 20 17, the German Federal Senate also promulgated relevant bills.
On the legal level of autonomous driving, China has never stopped exploring. In July last year, the Ministry of Industry and Information Technology announced that according to the revision plan of national standards and industry standards, relevant standardization technical organizations had completed the revision of three recommended national standards for the automobile industry, namely, Terms and Definitions of Advanced Driver Assistance System for Road Vehicles (ADAS), Performance Requirements and Test Methods of Blind Area Monitoring System for Road Vehicles (BSD) and Performance Requirements and Test Methods of Lane Keeping Assistance System for Passenger Cars (LKA), and now they only need to be further merged.
The laws on autonomous driving in many countries have been promulgated one after another, which undoubtedly brought dawn to autonomous driving and opened another important milestone of autonomous driving.
The scale of intelligent infrastructure such as vehicle scene measurement has expanded.
Autopilot is difficult to land, and it also faces the problem of safety: how to prove the high safety and reliability of autopilot system is a difficult problem faced by almost all autopilot players at present.
In fact, it proved to be very simple, nothing more than a lot of tests. Speaking with data, after all, practice is the only criterion for testing truth.
1. Testing of many real scenes
A week ago, Wenyuan Zhixing Weiruide and its joint venture Wenyuan Yuexing released the trial operation report of China's first L4 self-driving taxi Robo-Taxi. In the first month of trial operation in Wen Yuan Guangdong Bank, 8396 travel orders were completed from 20 19 12 0 1 to 3 1, * * *, and the safety accidents were zero. The average daily travel service was 270.8 times, the highest number of orders reached 438, and the total mileage of service orders reached 465,438+.
As a well-deserved leader in the field of autonomous driving, at this year's CES show, Waymo also announced that its driverless team has been driving on public roads? 2000? More than10,000 miles (32,654,380+0.8 million kilometers), it has carried more than 6,543,800 passengers, and now it has 6,543,800+0.5 million stable users every month, far ahead of its competitors.
In addition, Baidu, Cruise Ma Xiao Zhixing, Zhijia Technology and many other companies have also done a lot of work in autopilot testing, which will bring more and more real and reliable data for autopilot testing and accelerate the commercialization of autopilot.
Image source: Ma Xiao Zhixing Guan Wei
2. Establish a multi-site autopilot test site.
If you want more self-driving cars on the road, it is obviously not enough to rely solely on the efforts of enterprises, but also the support of local governments. Up to now, Beijing, Chongqing, Shanghai, Guangzhou, Wuhan, Shenzhen, Jiangsu and other places have issued relevant policies to liberalize road testing, which has contributed to the commercialization of autonomous driving.
According to incomplete statistics, up to now, nearly 300 road test licenses have been issued in China, including traditional car companies, new car-making forces, internet giants, technology companies, etc., all of which have obtained different test licenses. What's more, places such as Wuhan and Beijing have allowed self-driving vehicles to carry out manned and cargo tests. Undoubtedly, the promotion at the national level plays a decisive role in the accumulation of autonomous driving mileage. For self-driving vehicles, only through continuous testing can the reliability of data and the safety of driving be enhanced.
3. The simulation test technology tends to be mature.
Rand Corporation, a famous American think tank, once estimated that a L5-class self-driving vehicle needs 1 1 100 million miles on the road. This means that even the self-driving fleet with 65,438+000 test vehicles will take about 500 years to complete the test mileage of 65,438+065,438+0 billion miles, and the test will be conducted 7X24 hours a day. Therefore, it is obviously not enough to rely solely on the real vehicle road test. Therefore, virtual simulation test has become one of the important means to accumulate test mileage for self-driving cars.
Simply speaking, the simulation test is to simulate the road test environment through sensor simulation, vehicle dynamics simulation, advanced graphics processing, traffic flow simulation, digital simulation, road modeling and other technologies, and establish realistic mathematical models of static environment and dynamic traffic scenes, so that self-driving cars and algorithms can test driving in virtual traffic scenes, covering the scenes to the maximum extent, thus achieving test mileage that is difficult to achieve in real life in a short time. More importantly, it can solve many limitations in the actual road test, especially for some extreme scenes that are basically impossible to test on the actual road.
In recent years, with the entry of technology giants such as Google, NVIDIA, Baidu and Tencent, and the arrival of 5G communication technology, the problem of delay in receiving simulation test data has been greatly solved. In addition, a number of new players have emerged in this field, such as AAI, 5 1WORLD, Cognetta, Panosim and Parallel? Domain, RightHook and Latent that Waymo just acquired? Logic, etc. The simulation test technology is becoming more and more mature, and the test scenario is getting closer to the real scene.
Enterprises have entered the game and pulled up the progress bar of autonomous driving mass production.
The difficulty of self-driving mass production must be inseparable from practical factors: the production cost is too high, especially in the exploration stage, because no company can solve the problem of smart car industrialization alone. Because of the wide range and high technical requirements of autonomous driving, it is difficult for a single enterprise to form effective competitiveness. In order to complete this huge project as soon as possible, it is necessary to gather car companies, suppliers, the Internet and many technology companies to give full play to the strength of the industry, thus accelerating the implementation of related programs.
Image source: Huawei official website
Fortunately, now more and more enterprises have seen the business opportunities of autonomous driving and entered the field one after another. Not only traditional car companies, start-up technology companies and BAT have joined this battlefield, but more ICT (information and communication technology) giants, such as Huawei and Apple, have also entered the field of autonomous driving, laying out automobile products in the whole scene. With the full cooperation of enterprises, large-scale mass production and the improvement of business model, the cost problem must be developing towards a lower trend.
Of course, if you have to give an exact time for autonomous driving, many companies have given a clear time for mass production. 2020-2022 will be a big year for mass production of autonomous driving models.
At this year's conference of Hongqi H9, the company clearly stated that the new Hongqi will achieve L3-level automatic driving mass production in 2020 and L4-level mass production in 202 1 year. Geely and Guangzhou Automobile are also expected to achieve L3 autonomous driving mass production in 2020; In addition, GM previously released a production car Origin. According to GM's forecast, Origin will be produced by the end of 20021at the latest and in 2022. Volkswagen plans to launch L4 self-driving car on 202 1; Mercedes-Benz and BMW will launch L4 and L5 self-driving cars in 2020 and 20021year respectively. Volvo and Ford set the mass production time at 202 1.
In addition to traditional car companies, the new forces that build cars have also announced their own self-driving mass production time.
Xpeng Motors plans to achieve high-speed self-driving mass production in 2020, equipped with Tucki P7 model; Weimar Automobile will also realize the mass production and carrying of L3-class autonomous driving in 2020; Zero-run cars mean that before 2020, all models can be upgraded to L3 automatic driving function through software; Singularity Auto is expected to achieve L3 self-driving mass production in 2020, and plans to upgrade to L4 through OTA later; Tesla also said that it will achieve autonomous driving in 2020.
Summary: autonomous driving is getting closer and closer to us. This "newborn" who is still in its infancy will soon stand up on his own. I believe that with the implementation of the mass production schedule of car companies, the pace of commercialization of autonomous driving will be faster and faster. Perhaps, self-driving vehicles "flying into the homes of ordinary people" are not far away.
This article comes from car home, the author of the car manufacturer, and does not represent car home's position.