The recent report on disabling autonomous vehicle developers – that's what we're dealing with a Robo-Report card – has just been released by the California Motor Vehicle Authority. The columns and columns contained therein do not quite reveal the secrets of the very mysterious self-propelled vehicle industry. However, on its many pages it becomes clear that the Silicon Valley hype around Robocars may cool off, but progress is on the road to getting people off the wheel: the 48 autonomous vehicle developers who have tested their technology on public roads , have driven a total of 2.05 million kilometers in December 2017 and November 2018, up from 500,000 the previous year.
These reports show how many times each company's vehicle "broke away" from autonomous mode and was switched to the old-fashioned hand-to-wheel manual mode. For example, Waymo reported that one driver had to handle all 1
However, many standardized tests lack the reports. They are poor tools to understand how well this technology works and how each company's progress compares to competitors. In the space problem, where the reports are feared, there is a danger that engineers will be encouraged to give birth to their cars, to keep the detachment low and to stop the statistics. Any public information can be enlightening. And there is knowledge that can be gained from these thousands and thousands upon thousands of pages.
Before we dive in, just a few words about the reports themselves. California is the only state that requires it (it also requires developers to report all collisions in public). And to understand them, some big caveats have to be accepted:
- The reports are unscientific, as each company reports its data in different ways and provides various levels of detail and idiosyncratic explanations of what has ripped its vehicles out of self-propelled mode.
- They are full of indistinct language and have no context. This was the first year that companies used a standard data entry format, but not all seemed to follow the rules. And second, their language is still very vague. Waymo, for example, cites a number of expressions for "unwanted movements of the vehicle that were undesirable under the circumstances". Uber refers to a ton of disengagement operations as "preventive takeover or at the discretion of the operator". Do these situations describe similar situations? Who knows!
- The reports are of little use to anyone who wants to compare competing companies because these companies do not perform the same tests. Waymo performs most tests in simple suburbs. Zoox focuses on the complex city. They are better for tracking the progress of any outfit, but still not great, as these companies change over time as and where they are tested. And comparing the miles that each company has covered does not really tell you much about everyone's progress.
- Crucial: These reports refer only to public roads in California . So we do not know anything about Ford, whose tests are taking place in Detroit and Pittsburgh. And we do not see any data for Waymo's increasingly important testing program in Phoenix. (Aurora contains data from his testing activities in Pittsburgh, but without specifying the individual lift options.)
If we ignore the erroneous idea of using trigger couplings as a measure, we can still learn much from them reports. They also include information on how many kilometers the companies cover, how big their fleets are, and on which roads they are traveling.
Let's start our 21st-century Kremlology by pointing out the mix of actors. To classic cars such as Waymo, General Motors' Cruise, Nissan and Uber are still little-known newcomers like Roadstar.Ai, aiPod, Nullmax and WeRide added. Smaller names have become big names: Aurora has only raised $ 530 million in grants, Nuro has earned nearly $ 1 billion on delivery, and Apple has finally revealed some information about its secret (and potentially restless) self-propelled project. All three submitted their first withdrawal reports this year. Yes, the money in Autonoworld is not as free as it used to be. However, the growing competitive field shows that the dream of self-driving car is still alive.
The reports also show that almost all test farms of a company grow. Cruise tested 109 vehicles in the last reporting period. Now it has 194. (Many of these news are the generation 3, the company's electrical production model, AV.). Waymo increased the number of cars in this state almost by a factor of five. Uber is the exception: Almost a year ago, the self-drive program in California was closed after one of its test vehicles in Arizona had beaten and killed a woman. Overall, the DMV had agreed to 665 Robocars to test in California by the end of November 2018, compared to 326 the previous year.
Some of these vehicles built more ground than others – and show how each company's self-propelled strategy deviates from its competitors. Aurora's five vehicles only traveled 32,858 miles in autonomous mode, far less than Waymo or Cruise, with a higher retreat rate than these two companies. This small number is intentional, says Aurora co-founder Sterling Anderson. Ride around in public costs money: fuel, pay for the drivers sitting behind the wheel (and getting equity at startup) and the risk of falling. Aurora develops most of the development in simulation and on a test track. It reserves its on-road tests to publicly test how its technology is progressing. If there are no high rates of retreat, it moves into a new geographical area with new challenges.
"Testing on the road should not be development," says Anderson. "Tests on the road are more checks."
Competitors might call this Spin – Aurora's way of explaining low miles and long distances. Interestingly, the company is a second act to each of its three co-founders: Chris Urmson led the Google program (now Waymo), Drew Bagnell helped set up the Ubers team, and Anderson worked on Tesla's autopilot system. The fact that they re-established themselves in this unusual test method shows that the methods of developing this technology are evolving and diversifying themselves – a sign of a gradually evolving industry.
Other great WIRED stories