A Mercedes self-driving concept, back when they wanted to be fully in the race
The novel coronavirus crisis has slowed down progress in many fields, and in the world of self-driving cars, it doesn’t help that it came as people started worrying about a “winter” for technology growth after car companies didn’t make their made-up 2020 targets for deployment.
That “winter” made the earlier announcement that the planned partnership between BMW and Mercedes on self driving development was off seem a bit strange. After all, the car OEMs are happy to slow down at this point. They don’t want to be in a race with startups to upend their entire industry. They want it to happen at a pace they control. By declaring that there is not a big race, it makes perfect sense for car companies to combine their efforts. Mercedes’ rival isn’t BMW or most other car companies, it’s Waymo or the startup-like projects inside Ford and GM. Joining forces for a while saves resources which they started allocating to more near-term driver-assist projects to compete with their other real competitor: Tesla
Some of the reasoning was revealed in the declaration that Mercedes-Benz will put NVIDIA
NHTSA “AV-Test” data sharing plan does not make a lot of waves.
Looks fancy, but what does it really mean?
Last week, the US highway safety agency, NHTSA, announced their new framework for self-driving companies to report the results of their testing. It’s not a bad idea in general, but the reaction in the community was mild, because it only had a handful of companies as initial members and participation is voluntary.
The state of California has mandatory reporting on testing, and almost everybody tests in California, yet these reports have regularly come under fire from both media and the companies who have to report. The mandatory metric – disengagements – is both not very useful and bends what people to do so that they look good. That bending won’t happen with a voluntary report, and details on what will be in it are scant. Companies will report where they are testing, and what sort of vehicles as a minimum.
NHTSA touts there will finally be one place to find out all the testing that’s going on, but it’s not clear the lack of that place has been any particular glaring hole in the public’s knowledge, nor something that the press can’t do. The reality we don’t know how to measure safety yet, so making people report measurements we aren’t sure are right is not going to help anybody. We have to figure out how to actually measure the safety of testing and deployment, and then refine over time good metrics for it, and once we have those, you can make them mandatory if public safety will benefit from that.
Oxbotica hunting a buyer?
Oxbotica is a self-driving team spun out from Oxford. It was one of the more advanced early academic teams. Rumours have been reported about them seeking a buyer. This adds to the reports that Zoox is also looking for one (possibly Amazon) at a much higher level. Zoox has raised almost $1B while Oxbotica has raised only $30M. Still, this is another notch in the continuing consolidation expected around this time. While the world will support perhaps a score or more of self-driving companies, there are many more than that.
AEye moves to MEMS based LIDAR
The hunt for cheaper, more robust LIDAR continues. AEye (formerly US LADAR) has made “traditional” long-range 1550nm LIDARs but now has released a product with MEMS steering. Mems are tiny moving machines built in silicon chips, so small that people sometimes called them “solid state” though that’s not quite true. A million tiny mirrors reflect the light in a DLP home theatre projector, and your phone knows you shake it thanks to tiny MEMS devices. Aside from being cheap and small, MEMS steering can be more robust to vibration — a problem that has plagued many larger LIDAR designs. Several other companies also make MEMS based LIDAR, so we’ll see where the race goes. There are, it is reported over 100 different LIDAR companies, and they definitely can’t all survive either.
