Wednesday, March 28, 2018

On car sensors and Uber

Clarence emails:
Regarding the Uber car fatality, the excuse that it was dark is just that - an excuse. Mercedes-Benz has offered night vision on the S-class for ten years now. An emitter in each headlight cluster 'illuminates' the road ahead and a sensor at the top of the windshield receives the return infra-red waves. An image is generated for the driver on the instrument panel. BMW has offered their version of night vision for 5 or 6 years on the 7 series which gives the driver a warning when a warm body is detected.

Also why was there no LiDAR fitted to the Uber car? I think Uber is trying to do things on the cheap.
Yeah, this sounds right to me.  This is a big reason not to trust self-driving cars (companies skimping on the technology to the point that they don't really have it right).

7 comments:

Fredrick said...

Why is Uber doing this at all, they are a ride sharing service not an automotive manufacturer? Somebody should look into the business relationships between Uber executives and the manufacturers of sensors and software being used to see if they are just funnelling funds from Uber into a venture they are invested in to get the benefit when a software/sensor company gets bought out by venture capital.

Unknown said...

If I'm reading this article correctly, Uber disabled some of the on-board technology on the vehicle. And there are questions about how well the stuff that wasn't disabled was implemented.

"Intel Corp.’s Mobileye, which makes chips and sensors used in collision-avoidance systems and is a supplier to Aptiv, said Monday that it tested its own software after the crash by playing a video of the Uber incident on a television monitor. Mobileye said it was able to detect Herzberg one second before impact in its internal tests, despite the poor second-hand quality of the video relative to a direct connection to cameras equipped to the car."

Rick C said...

1 second? That doesn't seem like enough time to stop a car moving at 40mph.

Fredrick said...

@unknown,

There are plenty of open source videos available showing the same street in the similar evening conditions and they have no trouble seeing a pedestrian. If the visibility was that bad the "safety" driver should have taken control of the vehicle long before then.

Old NFO said...

Yep, they cheaped out... The Volvo did NOT have lidar.

Glen Filthie said...

If you wanted a self driving car with the most reliable sensor suites going - you'll end up with a car that will run you 6 figures and up - basically the price of a well appointed Cessna or private aircraft.

Self driving cars are a really, really BAD idea. Although, given the amount of idiots that have driving licenses and shouldn't - it might still reduce the body count...

Jonathan H said...

I see lots and lots of hype about self driving cars - and I don't see the technical rigor to back up that hype.
Most of the hype comes from internet companies who are used to shoddy products working well enough; they don't have the mindset of companies that deal with life safety equipment.
They should have the mindset of airplane manufacturers and autopilot manufacturers, who look at every little things that could go wrong and how to keep it from going wrong - and are willing to pay the price for equipment (both hardware and software) good enough to meet stringent goals.
Keep in mind that a self driving car that can deal with pedestrians and other traffic is a significantly tougher challenge than flying an airplane on autopilot - and it took 20 years of work to put out the first reliable autopilot!
Last year, self driving cars killed more people than commercial airplanes - that is NOT a good sign!