High control and sensor prices are two key factors that constrain the development of unmanned vehicles

The value of driverless development

The development of driverless technology not only liberates the human hands but also liberates the brain. Our attention does not need to focus on driving, which will bring great progress to economic and social benefits.

Undoubtedly driving in Beijing and the US Silicon Valley is not an enjoyment to us. It takes a lot of time and effort. If the driverless technology is popular, you can free up driving time. Check the mail, read the news, and even take a break. In addition, unmanned driving will also bring about changes in the ecology of the economy. For example, the unmanned industrial chain may change the location of the shopping district.

Data show that humans drive a fatal accident on the order of one millionth of a kilometer. Since the development of driverless driving, it has traveled tens of millions of kilometers and Uber has happened. Relatively speaking, it is more human than drone. Higher driving safety.

Driverless replacement of the driver is a long process and it can be seen that the driverless test is equipped with a driver. If driverless technology is popularized, it can be used as an overall smart city optimization. For example, everyone can have a centralized dispatching system to optimize the crowd at the same destination and promote shared travel.

Unmanned technology not only liberates human energy and efficiency, the ultimate direction is the establishment of a smart city, intelligent transportation planning. With the development of such an overall plan, the proportion of drivers may gradually decrease, but in the end, the driver`s day is replaced. There may still be a long way to go.

Domestic test environment is more challenging

Li Lixuan believes that the U.S. modern driverless technology is still far ahead of domestic ones. From the data reported by the California Traffic Management Bureau, it can be seen that there is still a gap between the domestic top Apollo and Google Waymo and Uber.

On the other hand, there are many unmanned talents in Silicon Valley in the United States. This is a very important difference. GoogleWaymo, Uber and so on have accumulated a lot of unmanned talents. Domestically, this is still at an early stage of accumulation.

The domestic testing environment is more challenging. The government has provided a lot of support. Both the traffic regulations and the technical facilities have given a lot of support. Plus, the Chinese have adopted new things quickly, like these mobile payments, O2O. All kinds of models have not been seen in the United States, and they are quickly accepted in China. There are great advantages in this regard.

Therefore, China's advantage in this kind of landing and transformation is very large. The United States is technically richer than China has accumulated. In addition to talents, as the Chinese talents gradually accumulate and erupt, the final gap will not be great.

Two major technologies to be solved

There are two parts of the technology that drone technology needs to solve most urgently:

The first is the balance between the capacity of the sensor and its value.

According to statistics of the authority of the French market analysis agency Yole Développement, smart driving mainly through the camera (long-range camera, surround camera and stereo camera) and radar (ultrasonic radar, millimeter-wave radar, laser radar) to achieve perception; the most advanced smart car adoption 17 sensors (only for autopilot applications) are expected to reach 29 sensors by 2030.

For example, laser radar technology is not "atomic bomb technology." This technology just requires more precipitation, more energy to do it better and more refined. Technically, there is the possibility of cost reduction.

Now every LiDAR manufacturer says that as long as it gives me a lot of money, I can do the cost down, so as long as the technical plan is set down, the cost reduction can certainly be reduced, and its more challenge is how to put this radar to precipitation More stable, more accurate, more suitable for the use of unmanned vehicles.

Second, driverless planning control

Driverless technology has been solved very well in normal driving, but encountered some anomalies, such as when some pedestrians fail to comply with regulations or extreme situations, how do we manage long-term problems through algorithms? This is a challenge.

Maybe the unmanned driver tests a few million kilometers before a scene of an Uber accident occurs. When unmanned vehicles are tested, they will try to avoid such things. In this area, planning control and simulators can be a powerful force. point.

Using simulators and artificial intelligence to detect the extreme capabilities of some vehicles, or the reaction of the vehicles in extreme cases, these scenarios are often not able to acquire data or learn and test by normal means.

Expect high expectations for artificial intelligence

Many people think that artificial intelligence is not smart enough. This is because people have high expectations for artificial intelligence. From the perspective of unmanned vehicles, the human brain is a neural network that has undergone many years of iterations, which means that you are born The time is a well-designed network. This network is called genetic and biological brains.

In addition, for example, if you start driving at the age of 16, actually your brain's perception has been trained for more than ten years. Your understanding of the world is not like taking a lot of pictures like the unmanned vehicle and training. The perception ability of the human brain is very strong, so artificial intelligence can really reach this person's perception ability and there is still a long way to go.

Artificial intelligence now has the obvious application in perception and prediction with the development of computer vision, but in the decision-making planning, the application is not so direct. With the development of artificial intelligence, decision-making planning has also begun to change in a data-driven direction.

We train our algorithm by collecting data on people driving and distinguishing it from the data on the drive of the machine. Let our algorithm drive more and more like human behavior. This is the direction in which artificial intelligence begins to infiltrate into decision-making planning. In the future, artificial intelligence will become a mainstream algorithm in decision-making planning.

The regulations on unmanned driving in various cities have just been introduced and are not yet sound, but this is also a good example of embracing the change in driverless technology. In addition, under the supervision of these laws and regulations, it is more legal and effective to improve the stability and capability of the entire system, and then make the system better.

Many people regard the development of driverless technology as a game between technology and law. Actually, this is more like a process of mutual development and adaptation.