The Wandering Earth 2: What unmanned driving technology brings to Tu Hengyu is not a disaster, but an exploration that reaches infinite perfection.

"The Wandering Earth 2" triggered great discussions and in-depth thinking about life and technology. Autonomous driving technology also needs to undergo soul torture.

Even though the movie is set in 2030, the highly automated driving (L4) car driven by Tu Hengyu unfortunately had a car accident, causing the death of his wife and daughter. This has caused people to have various concerns and suspicions about whether car autonomous driving can develop to fully autonomous driving (L5) level and whether it can truly withstand the test of reality.

(Picture source: Stills of "The Wandering Earth 2")

Can a perfect score on the autonomous driving test ensure that everything is foolproof?

People look forward to one day realizing driverless driving, but people’s worries and fears about driverless driving have never been eliminated. Wu Gansha, the founder of Uisee Technology who has been working on autonomous driving for 6 years as the director of Intel China Research Institute, made a metaphor respectively about the current autonomous driving technology and the maturity of autonomous driving: "If autonomous driving is... In a test, the current technology can score 98 or 99 points. This is the first metaphor. The second metaphor is that if the maturity and commercialization of autonomous driving technology are regarded as a marathon, it will probably run by now. One-third."

Wu Gansha explained: "We can understand autonomous driving technology as a test in an infinite question bank. However, in the unlimited question bank, you can get 99 points and get 99 points. It does not mean that you are close to 100 points. In this industry, we have a 90/10 principle. It seems that 90% of the road has been completed, but the remaining 10% of the road still requires 90% of the time and effort. In other words It is very difficult to improve from 99 points to 100 points."

"This is a qualitative analysis. If we measure it from a quantitative perspective, take the data of Waymo, the industry's leading autonomous driving technology company. An important indicator is called MPI (miles per intervention), which is the average mileage interval between autonomous driving being taken over by humans. Note that accidents will occur if humans do not take over.) The data for 2018 should be miles per 11,000 miles. After a takeover, the mileage increased to 13,000 miles in 2019, and the mileage in 2020 was particularly good, reaching 29,000 miles. This may be due to the epidemic and there are fewer cars on the road. However, in 2021, the mileage dropped to only 0 . There is one takeover every 80,000 miles.”

And what about human driving? "Data on American drivers show that insurance companies will issue one insurance policy every 250,000 miles. There will be one police incident every 500,000 miles. There will be one injury-causing accident every 1.5 million miles. There will be one fatal accident every 94 million miles."

"Therefore, if autonomous driving is to improve its performance by 20% compared to humans, it would take 110 million miles to have a fatal accident. To obtain such statistically significant features requires the accumulation of 11 billion miles of driving data . Cars that require self-driving technology have traveled a total of 11 billion miles. If no more than 100 people are killed, it will be considered that it has reached the level of one fatal accident after 110 million miles."

But you need to get such a certificate It proves that accumulating 11 billion miles is difficult for any autonomous driving technology company to achieve. Because if this company has 500 vehicles tested on the road at the same time, it will take 100 years to accumulate data on 10 billion miles of autonomous driving. Even if MPI meets the standards, it is still difficult to prove that your technology is safer than driving. Therefore, my second metaphor is that we have reached a score of 99, but if it is a marathon, we have only walked one-third of the distance.

"Pessimistic people are always right, and optimistic people can always succeed."

Even if the autonomous driving test with the unlimited question bank can score 99 points, it will never be able to score 100 points. Will it never be commercially available? Wu Gansha doesn't think so. He pointed out: "There are three ways to commercialize autonomous driving technology. The first is an unlimited question bank. Although you can't get 100 points in the test no matter what, you can have a teacher next to you to correct mistakes at any time. This is what we call L2 assisted driving." Because the responsibility lies with the driver, the teacher will correct the error when the autonomous driving makes a mistake. "

"The second way is to fix the question bank. It is the autonomous driving in the park. In this way, sufficient testing can be carried out to ensure that every time Even a driverless driver can score 100 points. This is also the scene in "The Wandering Earth 2" where driverless trucks transport ammunition and help build bases."

(Picture source: "The Wandering Earth 2" stills)

"The third route is in an open space, or an infinite question bank, but its running speed is relatively low and the vehicle size is relatively small, allowing It is possible for him to realize commercial use in patrol and delivery scenarios without taking 100 points."

Therefore, from the perspective of practicality and commercial use, the commercial use of the first unlimited question bank will have to wait ten years or more to reach L4 level or above. It may take twenty years to enter the mature stage. If it is the second and third situations, construction machinery in ports and docks has now reached the level of commercialization.

Who will win between bicycle intelligence and vehicle-road collaboration?

Currently, the route adopted by many autonomous driving companies is single-vehicle intelligence, while vehicle-road collaboration is another route that can quickly mature the commercialization of autonomous driving. The advantage of vehicle-road collaboration is to control the overall situation from the "God's perspective" and coordinate all vehicles on the road, rather than considering the "game" with surrounding vehicles from the perspective of a single vehicle. Vehicle-road collaborative use of roadside data can greatly simplify the algorithm of single-vehicle autonomous driving, and even reduce vehicle-side computing power requirements and equipment requirements, while truly realizing large-area unmanned driving to save economic costs.

However, Wu Gansha believes: “In the future, bicycle intelligence will be the mainstay and vehicle-road collaboration will be the supplement.”

“Because in vehicle-road collaboration, bicycle intelligence represents the lower limit of capabilities. , lacking bicycle intelligence, it is unrealistic to rely solely on vehicle-road collaboration. It is impossible for vehicle-road collaboration to completely cover every road in China, and cars will not be like our 4G and 5G, which can be used by everyone’s mobile phone. Secondly, It is also very difficult for all vehicles to take advantage of vehicle-road collaboration. Therefore, it is impossible for vehicle-road collaboration to achieve 100% coverage nationwide, and it is estimated that only 1% of vehicles will be able to use vehicle-road collaboration in the next 2-3 years. "

"Although vehicle-road collaboration cannot be fully covered, it can still be achieved in some specific areas, such as airports, terminals or specific highways. However, for a technology to be applied on a large scale, it must achieve closed loop: Who Who will pay for the operation? Customers must be aware and willing to pay. If you just do some pilot projects, there is no way to promote and apply them on a large scale."

"In addition, vehicle-road collaboration is smarter than bicycles. It is more difficult to determine responsibility when an accident occurs. If it is a single-vehicle intelligence, as long as the problem of the car is solved, all the problems will basically be solved. However, the number of links in the vehicle-road collaboration will increase significantly, and the identification of the cause of the error will become more complicated."

The real battlefield of autonomous driving technology is not ToC but ToB

Many people are now focusing on the ToC market, but the real value of autonomous driving lies in the ToB market. Wu Gansha said: "The self-driving technology that everyone can perceive now is used in passenger cars, taxis, and driverless buses to help people get from one point to another. However, a large number of applications of self-driving technology The place of strength lies in logistics."

"From the perspective of a city and an enterprise, the most important thing is to transport goods. In the city, trunk logistics supports the operation of the city. In the enterprise, from raw materials and products They all need the support of logistics, which is as important as capital flow and information flow. Most companies will ultimately deliver the physical products they manufacture/produce to customers."

"Now everyone is talking about the digital transformation of enterprises. Its essence is to solve the problem of human uncertainty and inefficiency. Human uncertainty includes that labor is becoming more and more expensive and scarce, and fewer and fewer people are willing to do repetitive work. Even when driving, the driver’s skills The level is also uneven, and different people have different learning abilities, and the time to master skills will vary. People's emotions will also be affected by various factors. And autonomous driving technology can very well avoid these uncertainties of people. Factors. Therefore, after digitization, we can have more comprehensive wisdom and more efficient planning to ensure that the entire process of enterprise production achieves the perfect state of Just in Time."

ToB The autonomous driving application market is vast but faces three major challenges Dilemma

Autonomous driving has broad market prospects in the ToB market. According to a report from Carnegie Mellon University, cars with intelligent driving functions can increase fuel economy by 10%. The higher the automation level, the higher the energy saving efficiency. More cost-effective: Intelligent driving is of great significance to scenarios with high labor costs, such as long-distance truck transportation, which can save labor costs of 60,000 to 150,000 yuan per vehicle per year.

However, the lack of standards, misallocation of funds, and lack of scenario-based solutions are the three major dilemmas facing the market. Wu Gansha pointed out: "First of all, there is a lack of standards. Without unified standards, good technology and bad technology compete together, and blindly compete at low prices, resulting in the phenomenon of bad coins driving out good coins in the market. . The lack of standards is also not conducive to customer selection and hinders companies from adopting new technology innovation."

"Second, mismatch of subsidy funds. At present, local governments encourage enterprises to innovate and use fiscal subsidies to enterprises. This is a good policy. However, the government is also worried that some autonomous driving companies will engage in fraud. Subsidy. It is better to adopt post-type subsidies and distribute subsidies to users who adopt new autonomous driving technologies than to directly distribute subsidies to providers of autonomous driving technologies. This will encourage more companies to try new technologies and use autonomous driving technologies. Promote their transformation and upgrading, and at the same time, it also plays a positive role in promoting the promotion of new technologies."

Finally, lack of scenario-based solutions. Wu Gansha believes: "Autonomous driving technology is not a product, but a part of the solution and a part of digital logistics. This requires the joint efforts of industry associations, leading companies, and autonomous driving technology providers to provide real-world solutions suitable for enterprises." Application scenario solutions provide practical and effective solutions for their logistics problems."

Even if Tu YY only has 2 minutes of autonomous consciousness, Tu Hengyu will do everything possible to let his daughter live in the digital world To have a complete life. Even if the autonomous driving technology that can score 99 points is an endless marathon, it will still allow countless rigorous and determined technology and engineering personnel to continue to explore and improve. Although it is not perfect, they are infinitely approaching perfection.

The above is the detailed content of The Wandering Earth 2: What unmanned driving technology brings to Tu Hengyu is not a disaster, but an exploration that reaches infinite perfection.. For more information, please follow other related articles on the PHP Chinese website!