Wednesday, the California Motor Vehicle Department issued the batch of latest companies' testing of self-driving vehicles in the state. The reports provide the public with the best information on the slow and frequent testing of opaque on autonomous vehicles on public roads. But because companies have a lot of things about which they report their data, the story that it tells can be misleading.
By law, all companies that are actively tested on self-driving cars on public roads in California are required to disclose the number of miles driven and the incidence of human drivers obliging to control their vehicles without driving (also known as "dismissal").
The DMV defines seizures as "deactivating the autonomous mode when autonomous technology relies or when the safe operation of the vehicle requires that the driver demands autonomous testing of the driver of autonomous modes and immediate hand management of the vehicle. "
This definition of definitions allowed AV companies to report certain events that may be eligible for how you can see it. Last year, for example, a car-light GM-owned auto-driving car moved to San Francisco after the safety driver had control over a brake. However, the incident did not include the incident in its report, as the human driver did not act as a safety or unresolved safety concerns of the autonomous system by Cruise performance.
Without definitions on when a safety driver should take extra and more information from each company about where and when the test is taking place, critics say that there is no comparison. The mile driven in Palo Alto is not the same as a thousand driven in San Francisco. Similarly, if a company is dedicated to highway driving, it should have much less unemployment. There may be a difference between time and hour ̵
This broad interpretation of the rule dismissed some of the unemployed descriptions together. "They are absolutely meaningless," said Sam Abuelsamid, a senior analyst at Navigant Research.
Others who see some value in the reports. "There is no other metrics, they are an interesting portal that are the importance of making general efforts and progress," said Michael Ramsey, a senior research director at Gartner.
If the race is a good interpretation of the two best performances: Waymo Alphabet and GM Cruise. The two thousand launched significantly more in 2018 compared to the previous year and lower unemployment rates.
Waymo drove 1.2 million miles in California in 2018, more than three thousand to 352,000 thousand companies in 2017. To be sure, Waymo drove less than a thousand in 2017 when Google's spinoff launched most of its fleet to Arizona in anticipation that its first commercial taxi service was launched. That service went live (type) at the end of 2018, and at the same time, Waymo has challenged his efforts in his native state. The company's reduced rate fell to 0.09 per 1,000 miles self-driven, or one against 11,017 miles.
Waymo says that his lower unemployment rate suggests that cars are getting better on handling "edge cases," or situations that human drivers could not encounter once or ever – in their lifetime. This will be useful as Waymo grows its fleet size and extends into new territories, the company said in a blog post. (Waymo did not show the number of vehicles he has in the California fleet in his report to the DMV.)
"The way to learn and improve is so that this technology can give a scale to our self-esteem vehicles strong and different types of experiences, and reveal many of those edge cases, as we can, "says Waymo.
GM Cruise reported a 447,621 mile drive in 2018, more than its third in 2017. The company reported 86 unemployment, or 5,205 thousand. This transfers to 0.19 unemployment per 1,000 miles. Cruise says that it operates 162 self-driving vehicles in California.
AV experts like the total number of miles driven independently, but the quality of thousands. Cruise vehicles operate mainly within the San Francisco city boundaries and on roads that may be more complex than the towns of Silicon Valley, such as Mountain View, Palo Alto, and Cupertino. "Based on our experience, every test minute is rewarded in San Francisco as a suburban test hour," said Kyle Vogt, chief technician at Cruise, in the Media post in 2017.
Sneaking into third is Apple's place, which reported 79,745 thousand and 6,951 unemployment, or about 1.15 thousand unemployment. That is a lot of interventions, and Apple is sure to get together compared to Waymo and GM Cruise. The report shows that Apple had 62 self-propelled cars on the road, but suggested other details that Apple is closer to the 70-car mark. In the past, Apple has criticized the reporting requirements, requiring the DMV to "amend or clarify its position" in relation to dismissal and testing without safety drivers ".
Apple is a tough time for Apple's self-driving program, "Tidy Project". "Last month, the FBI asked Apple's employee of a Chinese citizen to steal trade secrets related to the company's autonomous auto project, and the second time the Apple employee's government was cut off at the request of their own resolutions – Steaming in the last seven months. Apple recently started with around 200 employees from the project.
Zoox (0.50 unemployment per 1,000 miles) includes Nuro (0.97 unemployment per 1,000 miles), and Pony.ai (0.98 unemployment per 1,000 miles). Tesla continued his tradition of reporting one-mile drive in autonomous mode. (The electric car company said it "shadow-test" on autonomous technology by collecting data from its cars equipped by Autopilot already on the road during normal operations.)
In total, 48 companies reported driving 2 million miles in autonomous road on California roads and highways, the DMV said. But without a standard process or way to verify the accuracy of the data, it is not clear if the reports are making a great deal to improve the confidence of the community in technology.
The shows show that most people are still skeptical about self-driving cars and they do not have to succeed in giving up control. Greater transparency for testing would be a long way to improve the public's attitude to robot cars. But until the government is successful and it requires more from AV operators, these DMV descriptions are all in reality.