DMV's latest ranking: Google ranks first in autonomous driving, while Apple ranks second
▲Click above Leifeng.com Follow
Text | Dazhuang Travel
Report from Leiphone.com (leiphone-sz)
Today's autonomous driving industry is full of competitors, all of whom are constantly accumulating test mileage and expanding the size of their fleets. But it would be difficult to rank them, because the data that can be used to support them is so scarce, and there is no complete standard for judgment in the industry.
Fortunately, California requires companies to submit statistical data regularly, and these data have become the best window for us to understand the technical level of each autonomous driving company.
The most valuable data submitted by each autonomous driving company are the "disengagement rate", cumulative test mileage and fleet size. The latter two factors are easy to understand, while the former refers to the frequency with which autonomous driving test vehicles require human intervention.
Waymo ranked first, GM Cruise ranked second
*Number of self-driving vehicles registered in California
Judging from the data, Waymo, which has been in operation for ten years, is still far ahead of the rest, with a safety driver only needing to intervene once every 11,017 miles, while the average American driver only drives 14,000 miles per year.
Waymo noted that the disengagement rate for its test vehicles (a total of 111 vehicles) has dropped by 50% compared to 2017, and the average mileage between driver interventions has increased by 96%.
In 2018, Waymo's test mileage in California also surged to 1.2 million miles. In comparison, in 2017, Waymo test cars only ran 352,000 miles in California.
The DMV said that 48 of the 62 companies that have obtained California autonomous driving test licenses have submitted statistical data. These companies accumulated a total of 2.05 million miles of test mileage in 2018, 500,000 miles higher than in 2017. This means that Waymo, with 110 test vehicles, accounts for half of the test mileage in California.
Data released by the California Department of Motor Vehicles (DMV) showed that GM's Cruise ranked second (with a total of 162 test vehicles), with Cruise's test vehicles requiring safety driver intervention only once every 5,205 miles, a 321% improvement over 2017. Zoox, which also received a large amount of funding, ranked third on the list with a disengagement rate of once every 2,000 miles.
Compared with the two giants, Apple, which has always been mysterious, is simply weak, and the Titan project does not seem to be going smoothly.
According to Leifeng.com's New Intelligent Driving, Apple now has 62 test cars in California, which drove a total of 80,000 miles last year, but the safety driver had to intervene almost every mile (safety drivers are really tired), and the result was second to last (Uber was worse than Apple), which was really surprising.
Image source: https://m.pedaily.cn/news/440407
Chinese teams are following closely and catching up quickly, while other countries have clearly fallen behind the United States and China in terms of technological strength. China’s top players Baidu, Pony.ai, WeRide, and AutoX are all on the list.
Tesla
Tesla is even more bizarre. Even though it claims to be a leader in autonomous driving, it stated in its report that it has never tested or operated any vehicles in autonomous driving mode on California roads, so Tesla does not have a disengagement rate problem. Tesla's approach really makes people worry about Musk's "flag" (to achieve autonomous driving by the end of the year).
Regarding this situation, a Tesla spokesperson explained, "Tesla develops self-driving cars through simulations, labs, test tracks and roads around the world. At the same time, because Tesla is the only manufacturer with hundreds of thousands of test cars (testing autonomous driving technology in shadow mode), we can learn from billions of miles of road driving. Tesla analyzes the data of the fleet through OTA transmission technology to safely improve the performance of Autopilot and future autonomous driving systems."
Although this explanation is reasonable, it is indeed strange that Tesla deliberately avoided California. After all, Tesla started internal testing of HW3 last year. Could it be that the hundreds of employees involved in the project are not in California?
Aurora
Compared with other companies with large fleets, Aurora is small and beautiful. It has only five test vehicles and has accumulated 32,858 miles of test mileage. Such data is just a drop in the bucket compared to Waymo and Cruise, and Aurora's disengagement rate is not low.
However, Aurora co-founder Anderson said that the high disengagement rate was intentional. In their view, testing self-driving cars on the road is not only costly, but also likely to cause traffic accidents. Therefore, Aurora prefers simulation and field testing. When the disengagement rate decreases, they will increase the difficulty and continue to "make things difficult" for the test cars.
“How can road testing be considered R&D?” Anderson said. “Road testing is more like verification.”
Self-driving truck companies are also joining in the fun
Embark, a San Francisco self-driving truck startup, also submitted its annual report (according to California regulations, self-driving truck companies are not required to submit reports). Last year, its semi-trailer trucks completed 124,062 miles of testing, with a disengagement rate of once every 1,392 miles. Although the results are not as good as Waymo and Cruise, they are better than Apple and Aurora, and they are also the third manufacturer to exceed 100,000 miles of testing in 2018. Since its establishment, Embark has raised a total of US$47 million.
Embark CEO Alex Rodrigues also proudly stated: "The data released today shows that Embark's solution is a real cost-saving tool. At the same time, focusing our main efforts on highway conditions will also allow us to commercialize autonomous driving technology faster."
Is the DMV report the “autonomous driving bible”?
To be honest, the data released by the DMV is of limited value because, like many so-called standard tests, it is not comprehensive and cannot reflect the complexity of each company's testing environment (weather, speed, road conditions, lighting conditions and other factors will affect the vehicle's autonomous driving performance). It is not realistic to rely on this report to gain an in-depth understanding of the autonomous driving industry.
In addition, in this rapidly changing new industry, everyone knows that secrets are everything, but for the future of the company, it is inevitable that some companies will "tamper" with the data.
Therefore, before you are misled by the rankings in this report, please give yourself the following precautions:
1. The report is not scientific enough because each company reports data differently and the level of detail is not uniform, such as what exactly causes a vehicle to disengage from autonomous driving mode;
2. The data submitted by the companies contained many ambiguous descriptions and generally lacked context. Although the DMV provided a standard data sheet this year, the reports showed that not every company followed the requirements;
3. It is unfair to use the data released by the DMV to rank autonomous driving companies, because everyone's test projects are different. For example, Waymo prefers suburban roads, while Zoox prefers cities with complex road conditions, so the accumulated mileage sometimes cannot tell you anything useful;
4. Most importantly, the data submitted by the companies only includes road tests in California. Don't forget that although California is the well-deserved center of autonomous driving research and development, Arizona is actually the real stage for autonomous vehicle testing and early deployment. In addition, Ford, another giant, is based in Detroit and Pittsburgh, so they don't appear on the list at all.
In summary, based on the DMV data, we have no way to verify whether the companies are telling the truth. Without a standard process to verify their records, the credibility of the reports submitted by the companies is no different from the restaurant ratings on Meituan. This data is indeed useful, but it is not appropriate to use it to judge the progress of each company.
◆ ◆ ◆
Recommended Reading
Google Cloud’s new CEO made his first appearance. What is the “turnaround plan” he proposed?
Apple may hold a press conference next month, but the new products are not hardware
Tencent QQ denies the "212" incident; Qualcomm fined $180 million in South Korea; Google pays Apple nearly $9.5 billion in "tolls"
Trump urgently issues the "AI Initiative" executive order: everything is for the United States to dominate artificial intelligence
True 5G is still some time away, but ridiculous fake 5G marketing is already happening
Snapdragon 712 mobile platform is officially released, and Qualcomm has also started to squeeze toothpaste
Apple officially announced the departure of its No. 2 executive, who was once predicted to be the company's next CEO
The "2018 AI Adaptive Education Industry Research Report" was released on January 14, 2019. The current early bird discount price is ¥599, and the original price will be restored to ¥699 on January 20. It is a must-read for educational technology researchers, entrepreneurs, education practitioners, and investors. Scan the QR code for details.
京公网安备 11010802033920号