Four years ago, Baidu released the autopilot platform in the name of Apollo. The implication is obvious: promote autopilot landing like Apollo landing on the moon. At that time, Baidu groped for four years, and the technology has evolved to the third generation.
The real Apollo program began at 196 1, and Armstrong took a "small step" to land on the moon at 1969, which took eight years. Baidu has been engaged in autonomous driving since 20 13, and this year is exactly eight years.
On June 17, in Beijing Yizhuang Apollo Park, Baidu released the latest fifth-generation unmanned robot Robotaxi-ApolloMoon. The moral is also obvious: this "moon landing" plan was a success in one fell swoop.
Beautiful moral, as well as the metaphysical blessing of Apollo's "eight years of success", the atmosphere has always been full. The rest depends on the fate of Apollo moon and Baidu.
0 1
Apollo, the power of the moon
Summarize the official strength of Apollo moon with a set of figures: the production capacity is increased by 10 times compared with the previous generation, the delivery success rate is 99.99%, the cost price is 480,000, and the hardware reliability is 2000.
The industry knows that there are three obstacles to the landing of autonomous driving: technology, cost and regulations. Laws and regulations mainly rely on the government and industry organizations. What enterprises can do is mainly technology and cost.
The characters in Apollo Moon's book are aimed at these two points.
On the technical level, the sensing hardware of Apollo moon is roughly similar to the fourth generation products that are being put into trial operation. Jose's lidar (2) is standard, which increases the number of cameras and improves the resolution and frame rate, but in general, they are limited. Nowadays, autonomous driving has passed the stage of stacking hardware, and enough is enough.
The official capacity is 10 times that of the previous generation, which is a bit vague. It should mainly refer to computing power, algorithm and overall system efficiency. The platform computing power of 800tops is very powerful (higher than this, it is said that the chip computing power will surpass Weilai ET7 and Zhiji L7); 1000 tops)； The optimization of algorithm and system efficiency is self-evident. The open road test has been carried out for nearly two years, and the test mileage exceeds 654.38+0.2 million kilometers. The value is here.
In order to strengthen the memory, the official threw out four 9s (99.99%), emphasizing that Apollo's landing on the moon was absolutely foolproof in the real scene. However, the test conditions and statistical methods are unknown, so it is not easy to evaluate for the time being.
Compared with technology, Baidu is more willing to talk about cost this time. Not only did it generously show the cost price of Apollo landing on the moon of 480,000 yuan, by the way, the average cost of Robotaxi in the industry is about three times that of Apollo landing on the moon. This move is quite fatal. It not only creates motivation for itself, but also exposes the bottom of peers. In the future, there is expected to be an explanation for the cost problem.
According to Baidu's calculation, the cost is 480,000 yuan, and the monthly cost is about 8,000 yuan after five years of operation, which is almost the personnel cost of B2C network taxis in first-tier cities. The conclusion is that ApolloMoon has touched the cost threshold of online car. However, if the salary of the security officer is included, it is still relatively high.
Without considering reliability, there is no cost. The first three generations of Robotaxi on Apollo platform are all ready-made products with autopilot hardware, which is very strange and has a high failure rate. The fourth generation cooperated with Hongqi, pre-installed and mass-produced, and the reliability began to have a spectrum. On the Apollo moon in cooperation with polar fox, the hardware reliability is said to have exceeded 20,000 hours, and it has been running for 5 years without worry.
02
The difficulty of landing on the moon. "
Obviously, on the Apollo moon landing, Baidu's strategy changed. Instead of simply emphasizing the "good" of its own Robotaxis, it clearly told the outside world that Robotaxis can no longer be trapped in the struggle between "good" and "expensive" and can land on a large scale.
According to Apollo's early planning, Robotaxi will grow to 30 cities in 2023, with 3,000 cars and 3 million registered users, and will be partially commercialized.
At present, the test fleet is 500 vehicles, and the hope of quadrupling or quadrupling depends largely on Apollo landing on the moon. To this end, the government has set a radical iterative goal: upgrading products in two years, increasing production capacity tenfold and reducing costs by half.
Technology iteration and cost reduction are not difficult, if not "ten times" and "half". Similarly, with the determination and ability to continue to burn money, it is not difficult for the Robotaxi team to expand to 3,000 vehicles.
The problem is that even if everything goes as expected, Baidu Apollo may not be able to successfully "land on the moon."
Waymo, which has gone the farthest before, has solved the problems of poor Robotaxi experience, low acceptance and, most importantly, insufficient security (especially easy to be rear-ended) after nearly three years of trial operation.
The operation report of Apollo test team last year was quite gratifying: 2 1 10,000 person-times, with an average daily order of 2,703 in a single city, and the overall customer satisfaction exceeded 95%. But obviously, the result had to be discounted. First, the scene of trial operation is limited, and second, most young people are early adopters. For example, in Beijing, the most popular route is between two subway stations. Both office workers and student parties are experiencing fishing. When they are finished, taking the subway back has little to do with the real scene.
What percentage of freshness, acceptance and satisfaction can be maintained is uncertain. In reality, it takes courage to make the machine safe. This kind of courage to push the current technical road to the limit cannot be given. After all, no matter how advanced the algorithm is, it is impossible to exhaust all the small probability scenarios. Machine decision-making is impossible for anyone to imagine.
The 99.99% delivery rate of Apollo landing on the moon is true, but there is only a difference between 0 and 1 in safety. The probability of 0.0 1% can also prevent Robotaxi from coming to the door. As for the failure prevention measures such as 5G cloud cost and V2X, the bottom layer is OK, but it can't solve the security problem between cracks.
Apart from the technical restrictions that are temporarily insurmountable, the landing of laws and regulations is temporarily unknown. Although all localities are scrambling to build unmanned demonstration zones, they are busy issuing test licenses. This is called taking the lead in development. It is another matter to make rules and let Robotaxi land on a large scale. There is no timetable at present.
03
Baidu's anxiety
In this case, the autopilot competition is about who can do it in the short term and who can dress well in the long term.
The problem is that Baidu may not be able to afford it. Eight years of autonomous driving research, Apollo4 has been launched for four years, with tens of billions of investments, 150 partners and nearly 3,000 autonomous driving patents. So far, there is no decent harvest. More importantly, we can't see where the predictable benefits are.
Robotaxi, as the ultimate picture of autonomous landing, is of course beautiful, but just like Apollo landing on the moon, it takes time and energy and depends on opportunities.
Initially, Baidu launched the Apollo open platform, hoping to attract more car companies and suppliers to join in and use their own solutions, so that the income can at least partially support autonomous driving projects. But when it comes to the right to speak in the future, car companies are playing Tai Chi, smiling at Apollo, undertaking some cooperation projects that have nothing to do with the overall situation, and turning around and continuing to use the mature L2 of Tier 1. Have the strength to engage in autonomous driving.
In desperation, Baidu pulled Geely and established Qiqi. The next time it builds a car, it will add some chips to Apollo's landing. After the release of Q 1 financial report, Li Yanhong clearly stated in the letter to all employees that the commercialization of Apollo should be "three legs": providing solutions for original equipment manufacturers, building their own cars, and * * * enjoying Robotaxi.
However, as we all know, the latter two items have no profit prospects in the short term, and only Xiu Xiu's strength helps Apollosolutions to open the market on the OEM side.
Stripping off Apollo business and financing independently may alleviate Baidu's anxiety to some extent.
Coincidentally, when Apollo landed on the moon, Vaimo, who experienced the recent personnel turmoil, announced that it would raise 2.5 billion US dollars, plus 3.25 billion last year, and accumulated external financing of 5.75 billion.
Almost at the same time, Cruise, a subsidiary of General Motors, also announced that it had obtained a 5 billion credit from General Finance for the mass production of the driverless car Origin. Previously, Cruise has received more than $654.38 billion from external investors such as Softbank and Wal-Mart, and has enough grain and pasture.
Seeing that the investment in autonomous driving is heating up and colleagues at home and abroad are enriching blood, Baidu can't have no idea. First, I don't want children who put their hopes on themselves to eat "a hundred meals"; Second, Apollo's external valuation may never meet Baidu's expectations.
Last February, UBS valued Apollo at $654.38+000 billion. In contrast, although Waymo has fallen by more than half from its peak, it still has 30 billion pounds, and Cruise also has 30 billion pounds. Apollo of the first battalion was obviously not satisfied. Perhaps, after repeated actions this year, the capital market will have different evaluations.
In the original Apollo moon landing plan, Apollo 1 was destroyed by fire. Before Apollo 7, the orbit around the earth was unmanned. Apollo 8 entered the lunar orbit for the first time, and both Apollo 9 and Apollo 10 were carrying out tests around the moon. Finally, the historical moment of Apollo 1 1 has arrived.
Baidu's Apollo obviously can't stand such a toss.
The above content was uploaded and published by "Cheba Ke", check the original text.
Million car purchase subsidy