The data as the foundation,the technology as the bridge,How far is the era of driverless?

Recently, domestic companies have been exploring and practicing more and more in the field of unmanned driving. Following Baidu’s announcement of the mass production plan for the red flag autopilot model, GAC’s auto-driving taxi appeared on the streets of Guangzhou, and Debon’s unmanned trucks completed the first large-size express delivery. For a time, many companies launched a new round of impact on unmanned technology.

Autopilot or unmanned technology seems to be still far away, but in the course of continuous practice, there has been a large degree of development. In the process of gradually landing, the main players in the field have completed the layout in the basic technology field, then which technologies and applications are the focus of their competition? Which step has the entire autonomous driving industry in China taken to the ground? In the future, how will autonomous driving change the business form of related fields?

Unmanned driving requires basic technical support BAT basic layout

The technology involved in driverless driving is complex and complex, and here are some of the more popular and important aspects. Of course, the most important thing is the car itself. If you want to drive automatically, you must either build a car or you need to cooperate with the car company. Professional people do professional things. For the Internet companies like BAT, cooperation with car companies is a common solution. The data they want, many of them come from the original factory, the data of the two parts of foreign auto parts are obtained through channels or cracks.

In addition to the car itself, there are several key technologies that have a huge impact on autonomous driving capabilities, and BAT is not far behind in these areas.

The first is the collection and basic perception of the external environment, where high-precision maps are an important source of information for autonomous driving. Traditional maps are only used as information for drivers, and high-precision maps are used for AI to make driving decisions directly. Therefore, this piece is also an early part of the autopilot company's efforts. Baidu map (Changdi Wanfang), Gaode map and Tencent's four-dimensional map are all BAT layout in the field of high-precision maps.

In addition to high-precision maps, various types of sensors are also essential for cycling intelligence. In the period when the road coordination is still underdeveloped, the automatic driving relies more on sensors such as cameras, laser radars, automotive CAN bus data, and electronic steering. These technologies are also highly demanding and costly. At present, there are some achievements in the domestic laser radar, such as Sagitar and Hesai Technology, which have already been invested in Ali and Baidu respectively.

The second is computer vision or image recognition, which is to analyze and process the collected information. This is a complicated process. Especially for ever-changing road conditions, it is difficult to be accurate enough. At present, Baidu Brain and Tencent U-Lab are all tackling visual recognition technology. Ali's Shangtang Technology is also exploring the application of its more mature recognition technology in the field of automatic driving.

The third is the information exchange between the car and the outside world, that is, V2X, which can be combined with technologies such as intelligent roads and cloud computing, and is an important part of the vehicle road coordination scheme. By giving more information to the road to collect information, cars use V2X technology to share information with smart roads and other road users in the Internet of Vehicles, which is relatively more accurate and efficient. At present, Baidu and Ali's two programs are the most concerned, and the network of car and time-share leasing companies also joined the team.

The fourth is the in-vehicle OS and AI chips, which is the only way to achieve other technical applications. BAT has its own in-vehicle system and also carries its own ecological goals. The differences in OS are beginning to show a tendency to close, and different ecosystems are separated. This will have some inconvenience for the communication of information in the future Internet of Vehicles. Smart traffic will inevitably require data to be opened. At that time, OS with higher market share may grasp more. More voice.

As for the AI chip, the future is definitely an indispensable part. In addition to the BAT has been released or is being researched and developed, there are many companies entering the market, among which there are many hardware manufacturers such as Huawei, Su Rui, and Hui Wei. But whether it can meet the needs of autonomous driving, and ultimately depends on the effect of landing.

Thread has solved the high-speed video capture and decompression by using I.MX6. The Qualcomm 4G module will transmit the CAN bus data collected by Freescale S32K and the peripheral GPS and other SENSOR to the server to solve the problem from the acquisition end to the decision side. Data problem, build driving model by establishing vehicle CAN bus data such as vehicle speed, steering angle and gear position information, and establish samples.

The difficulty is being gradually broken. Which step has the autopilot been taken?

The basic technology has been gradually promoted and completed, and the speed of development of autonomous driving has also accelerated significantly. This is also the reason why new and recent progress in this field has been continuously reported this year. Let us observe from the technical maturity of autonomous driving itself and the level of infrastructure support it needs, and at what stage does autonomous driving go.

(1) Has developed to the L4 stage, and has basically achieved autonomous driving in terms of technology.

The level of automatic driving is divided into 5 levels, which are L1~L5 according to the degree of automation from low to high. For the classification criteria, the definitions of NSHTA and SAE are slightly different. But basically it can be assumed that L3 is conditional automation and requires human intervention; L4 is highly automated, and manual intervention is required only in special cases; and L5 is fully automated and requires no manual operation.

At present, Waymo, Baidu and other unmanned vehicles have entered the L4 stage, and the auto-driving taxis tested by Wenyuan Zhixing in Guangzhou are also L4. This stage can basically meet the normal driving needs, which is already an ideal state of automated driving.

(2) 5G technology is gradually advancing and it is expected to mature in the next few years.

Second, the network foundation is gradually available. Autopilot technology has high requirements on the network, requires a large amount of information, and guarantees real-time transmission rate and stability. This requirement means that there is a certain risk in the large-scale use of driverless technology under the current 4G network, so it must rely on the maturity of 5G networks.

Fortunately, the formulation of the 5G standard has been completed, and now all countries are investing in the landing of 5G networks. The United States, South Korea and other places are ready to officially commercialize 5G networks next year. The three major operators in China also indicated that they will build 5G network commercials in the third quarter of next year, and more than ten first- and second-tier cities have already tested or built base stations. When the autopilot technology is put into use in the future, the maturity of 5G can also be a heavy responsibility.

(3) Car routing has made great progress. Baidu wants to open source Ali to pay attention to freight.

 Finally, the road coordination program, which can greatly improve the efficiency of autonomous driving applications, has also made great progress. Ali has already tested the open road section in Hangzhou. Baidu will officially open up the Apollo road-to-road collaboration program at the end of this year to open the technology and services of Baidu Apollo in the road-to-road collaboration field.

It can be seen that Baidu seems to be aware of the trend of data interoperability in the future. There will also be standard disputes in the road-to-road coordination scheme. I wonder if Ali will also intentionally make corresponding actions in the competition for standard rights. From the current situation, when Ali proposed the vehicle-to-road coordination program, he set his sights more on the freight transportation field. After all, this is a more concerned part of Ali.

After the successful marketization, automatic driving will have a huge impact on all walks of life.

As more and more breakthroughs are made in autonomous driving technology, we are gradually seeing the rudiment of future business changes affected by it. Although we don't know when this day will come, we can predict that some future industry directions are likely to follow these trends.

(1) Multi-factor boosting Freight is most likely to be the first to achieve automatic driving

In terms of unmanned landing, freight is most likely to be the first to achieve this grand event.

First, e-commerce and logistics companies have a clear goal driven by commercial interests and are more willing to promote unmanned driving. Unmanned driving can save a lot of money on personnel costs, and there is no more than 4 hours of fatigue driving concerns. It can run continuously for 24 hours (provided that after the automatic driving technology has been effectively broken at night), it is still sharing electric electric logistics vehicles. Mainly, Ali has laid out the e-commerce + electric city distribution + sharing mode.

Second, the safety of e-commerce and freight transportation is less prominent and more acceptable to the public. If it is a passenger car driving automatically, people's concerns will take some time to eliminate. Third, the marginal cost of the road coordination program is lower. The driving route of the passenger car varies from person to person and from time to time. If it cannot be spread out in a short period of time, it actually plays a small role. However, if the fixed logistics route is modified, the marginal cost is greatly reduced due to the high usage rate.

 (2) Deeply change the travel industry, the network car will be challenged by the shared car

Since the automatic driving has a lower potential requirement for the driver's driving level, the public without a car (but with a driver's license) can fully use the shared car without the need to play the car, which is the main reason why the Drip is distributed to share the car. The market for online cars will be captured by shared cars, and shared cars will shift from a leasing model to a model that includes both leasing and a real shared economy. The shared autopilot network will be the ultimate form of human travel.

The reason why there is a real shared economic element is that, for a car owner, the car is idle, it is better to let it go out and make money. On the other hand, it can also liberate the network car operators and let them directly transform into shared car owners. From a social perspective, it also enhances the safety of people's travel.

At the same time, the number of users of the network car is shrinking, but there is still a market in which it exists. The service group of the network car is more concentrated on some passengers who have no driving knowledge and cannot respond in an emergency. The network car will be equipped with a driver, but the driver's labor intensity will drop.

 (3) Master more data resources

With the deepening of the car's intelligence, car companies are no longer simply manufacturing industries. Due to the application of sensors, V2X, automotive CAN bus and other technologies, as well as the control of the underlying system of the car, Internet companies need to rely on the relevant interfaces of the car companies, and car companies can better grasp the effective big data resources

In addition, the willingness of car companies to participate in travel will also be strengthened. Now, Geely and other car companies have already entered the field of travel. In the era of auto-driving, which is less dependent on drivers, car companies can concentrate on their own advantages. There are loose cost costs for protection, better publicity results, and more and more car companies participating in shared or networked vehicles.

In short, although the driverless technology can't be done overnight, it is definitely getting closer and closer to us. As the basic technology gradually completes the layout, the unmanned technology reaches the L4 standard, and the infrastructure will become more complete in the future. Autopilot will no longer be whimsical. As for who will be able to grab the big cake of this sunrise industry, it depends on who will break through in the actual application.