As a car editor at a technology media company, the word “Meta Universe” frequently appeared in our company once made Xiao Lei feel bewildered. For this reason, Xiao Lei decided to ask “Du Niang” directly. Baidu Baike’s description of metaverse is like this: the term metaverse was born in the science fiction novel “Avalanche” in 1992. The novel depicts a huge virtual reality. real world.
If you say that, Xiao Lei can understand it in seconds. Although Xiao Lei has not read the famous science fiction novel “Avalanche”, but the classic American Hollywood science fiction movie “The Matrix” and the “Number One Player” released in 2018 “Aren’t they all talking about the same thing.”
After probably understanding the concept of the so-called “meta universe”, based on professional instinct, the editor of Xiao Lei Motors began to replenish its application in the automotive field. However, before Xiao Lei, who suffers from the late stage of procrastination, draws up a content outline, Nvidia is the first to give its understanding of the application of “meta universe” in the automotive field.
Nvidia, as a high-frequency graphics company that often appears in the mouths of gamers, understands the meta-universe as a virtual game. In the eyes of Nvidia, everything in the meta universe can be virtualized, including the virtual creation of another digital “high imitation” earth that can be used to test autonomous driving technology.
Can Nvidia’s simulated earth bring hope to autonomous driving?
As early as 2016, the RAND Think Tank pointed out that in order to ensure safety, an autonomous driving system needs to be tested at least 11 billion miles (about 18 billion kilometers) to achieve mass production application conditions. This test distance is equivalent to 44 around the earth. Ten thousand circles .
Data source: Head Leopard Research Institute
You know, Baidu Apollo, a giant in the field of autonomous driving in my country, will have a test mileage of only 1.12 million kilometers in 2020. Obviously, whether it is from the aspect of test cost or time, the 11 billion miles of test data in real life is a task that is difficult to complete in a lifetime.
In fact, there are only two key factors that affect the task of road testing. One is the number of test vehicles; the other is the mileage accumulated by each test vehicle. Therefore, the way to complete the road test task is also very simple, that is, to increase the base number of test vehicles and allow these test vehicles to continuously brush their test mileage on the road. In fact, Baidu Apollo’s 112km road test data in 2020 was measured using 55 test vehicles.
As we all know, the current mainstream technology route for autonomous driving is realized by equipping lidar, and lidar is extremely expensive, and an automated driving test vehicle often needs to be equipped with several lidars. Therefore, the high cost of test vehicles per vehicle has become a roadblock for autonomous driving companies to expand the base of test vehicles, and can even directly bring down most companies that develop autonomous driving technologies.
On November 9, Xiao Lei saw the hope of a blowout of autonomous driving technology at the GTC conference held by Nvidia, because Nvidia used Omniverse technology to replicate a digital twin earth. On this twin earth, it can simulate buildings, cars, robots, physical phenomena, biological systems, and 5G edges, and it can also simulate a twin self like NVIDIA founder Huang Renxun during the development of the conference.
When the meta-universe corresponds to the digital twin earth, Xiao Lei Nao made up a lot of possibilities for the development of autonomous driving technology. For example, whether automated driving tested in a virtual space can solve high hardware costs and the most critical safety issues.
Meta universe will reduce the cost of autonomous driving testing
In fact, the most aggressive research and development of autonomous driving technology is not car companies, but RoboTaxi, which focuses on L4 and L5 autonomous driving. They are often the companies that spend the most money in the field of autonomous driving. In addition to their money spent on research and development, the cost of autonomous driving hardware is also a huge expense that cannot be ignored.
From the current point of view, several first-tier RoboTaxi companies, including Waymo, Baidu Apollo, and AutoX, already have the strength to put unmanned driving technology on the road for trial operation. For them, in addition to continuing to upgrade their technology in the future, what bothers them is the huge cost of autonomous driving hardware.
Take Baidu Apollo, the main RoboTaxi low-cost solution, as an example. Its test car Apollo Moon costs only 480,000 yuan. If Baidu wants to run 18 billion kilometers of road test data in 10 years, it will cost at least 172.8 billion yuan to deploy 360,000 test cars based on its current average test car running 50,000 kilometers a year. Such expenditure and layout are an astronomical figure for any company.
Baidu Apollo is also a company that focuses on low-cost RoboTaxi solutions. If it is replaced by AutoX, which is known for its stockpiling, the cost of completing 18 billion kilometers of road test is even more unimaginable. You know, according to industry insiders, the cost of AutoX Gen5 is more than three times that of Baidu Apllo Moon.
If the hardware cost of autonomous driving technology cannot be solved, RoboTaxi will not be able to expand the test vehicle, and the chain reaction will be that it will not be able to obtain more road test data, and it will not be able to complete the above-mentioned 11 billion miles test target. , The vision of autonomous driving is naturally difficult to achieve.
Obviously, if RoboTaxi is testing autonomous driving in the digital twin earth, it can completely purchase virtual lidar, millimeter wave radar and other hardware, and does not need to use physical test vehicles to conduct road tests. This is similar to us. There is a reason why you don’t really need to buy a car when playing “Need for Speed”.
On the other hand, in the time limit, each of us has a fixed 24 hours a day. In the digital twin earth, as long as it can ensure the accuracy of the test results, RoboTaxi can completely drive its virtual test car to “accelerate” the suspension, which can make the test more efficient without affecting the test results.
In fact, it is not unprecedented in the field of autonomous driving to obtain a large amount of road test data by turning on the “acceleration” hook. You know, Google ‘s RonoTaxi company Waymo once completed 15 billion miles of simulated drive tests by inputting 10 million miles of real drive test data.
Facts have proved that simulation testing has become the only way for the development and implementation of autonomous driving. Therefore, Xiao Lei believes that testing autonomous driving technology in a meta-universe digital twin earth similar to the one proposed by NVIDIA can not only greatly reduce the test cost of autonomous driving, but also improve the test efficiency of autonomous driving.
Meta universe will reduce the risk of autonomous driving tests
In Xiao Lei’s view, the reliability of autonomous driving technology is also the main reason why the RAND think tank put forward the 11 billion miles of road test standards, and the institutions that formulate relevant policies are also cautious about the attitude of autonomous driving on the road. You know, human life is the most precious at any time, and autonomous driving essentially entrusts human life to the machine. Therefore, for safety reasons, even so far, no more than 3 companies in my country have obtained RoboTaxi test licenses for unmanned driving.
Xiao Lei believes that since the country is cautiously issuing autonomous driving test licenses for safety in real life, in the meta-universe digital twin Earth designed by NVIDIA, the test permissions for autonomous driving are likely to be further relaxed.
On the one hand, the original intention of the birth of the meta-universe stems from the human imagination of future technology. We can regard it as an experimental field for cultivating future technology products. On the other hand, since the digital twin earth is a virtual space, it will naturally not threaten the lives of human beings in the real world.
Of course, since the digital twin earth is a replica of the earth, the rules and regulations on the earth will of course be copied to it. Therefore, in Xiaolei’s view, even if the autonomous driving technology is tested on a “fake” earth, the autopilot technology supplier should follow all the rules of the “real” earth as much as possible.
You know, for them, this is not a game, but a cutting-edge technology that will be applied in real life in the future and will carry life. If the automated driving test does not threaten people’s lives, then its research and development threshold will also be greatly reduced.
In the virtual world, because RoboTaxi companies only need to focus on R&D technology, they neither need to bear the actual hardware costs, nor do they have to worry too much about the safety of life brought about by autonomous driving tests. If the cost of testing and safety issues can be resolved, then a large-scale road test of autonomous driving technology is just around the corner. At that time, we could at least experience the technological charm brought by autonomous driving technology in that world in advance.
Although testing autonomous driving technology in Metaverse will neither endanger the safety of people in the real world, but also greatly save R&D and testing expenses, Metaverse is essentially a virtual world known for its code, and code is an algorithm. , All algorithms have laws.
Based on Xiao Lei’s understanding of the meta universe, Xiao Lei believes that the meta universe is like our online dating when we were young. She looks particularly real, but it is extremely illusory. However, this is like online dating couples meeting sooner or later, when they return to reality, they will still be full of unforeseen accidents.
Autopilot technology vendors in building up the digital universe through metadata twin earth when the research made about the autopilot as well. Although Metaverse may be able to accelerate the development of autonomous driving technology, this does not mean that the autonomous driving developed in Metaverse must be applicable to our current world, because changes in the real world are always higher than predicted.
In fact, even if Waymo has run 15 billion miles of road tests in simulation tests, it cannot guarantee that its autonomous driving system will be foolproof in real life.
There is an old saying in our country called “Everything is difficult at the beginning.” Autonomous driving is one of the greatest technologies in human history. It will certainly have to overcome many difficulties before it is truly realized, but we can also see with optimism that with the help of many new technologies, we have been able to see the hope of technology landing.
Posted by:CoinYuppie，Reprinted with attribution to:https://coinyuppie.com/metaverse-assists-autopilot-to-accelerate-landing-really-no-joke/ Coinyuppie is an open information publishing platform, all information provided is not related to the views and positions of coinyuppie, and does not constitute any investment and financial advice. Users are expected to carefully screen and prevent risks.