On Mon, 26 Sep 2016 12:26:00 +0100, Roland Perry
wrote:

In message
-sept
ember.org, at 08:45:29 on Mon, 26 Sep 2016, Recliner
remarked:
Roland Perry wrote:
In message
-sept
ember.org, at 21:30:40 on Sat, 24 Sep 2016, Recliner
remarked:

Anyway, here's a recent report of Uber's self-driving tests in Pittsburgh:

http://www.economist.com/news/busine...nches-its-firs
t-self-driving-cars-pitt-stop

I think you'll find that's the University's testing, and because Uber
funds that programme they get to go "along for the ride" so to speak.

It's also an early testing phase, which the cars won't necessarily pass.

It's not really a pass or fail issue. It's an alpha test. I assume the
software, algorithms and mapping database will be continually adjusted
during this testing phase, but no-one is planning to roll out this version
as a commercial release. But these improvements will feed back into the
eventual commercial release, which is probably several years away.

Like fusion power, you mean?

No, exactly the opposite. This stuff works, and just needs
fine-tuning. Fusion has never got that far.


The big step in this phase is that it's not just the private test running
that Google has been doing for years, but a public test, with random
members of the public actually using the cars as a taxi service.

I've always assumed the Google test was at the very least assisting
their employees to commute to work. Or is it only people driving around
at random during their work day with the firm?

I think it's just a test programme, with professional testers driving
around.


It seems to be a little more ambitious than the nuTonomy trial that
started a few days earlier in Singapore, but is still well short of a
commercial release.

The novelty is the way they are spinning it for PR purposes. I don't
blame them for that, but it does appear to have got many people rather
over-excited. Didn't BR run the APT in Alpha-testing revenue service,
before scrapping the project?

The self-driving cars are already much further ahead than the APT
reached.


As an aside, it's interesting how much much hardware these early
self-driving cars currently need (numerous sensors, Lidar, Radar, cameras,
etc) compared to just the eyes and ears we human drivers get by with.

20 cameras, and radar! I wonder how fault-tolerant it is when one or
more of the cameras fails because of an electrical fault, or become
covered in snow. And don't forget, one of the problems Volvo has found
is the radar antennas getting clogged with snow.

All stuff to be evaluated during the test phase. Hopefully the
commercial version won't need quite so many cameras and sensors.


ObRail: Right or wrong sort of snow, I wonder?

In message , at 12:46:36 on
Mon, 26 Sep 2016, Recliner remarked:

But these improvements will feed back into the
eventual commercial release, which is probably several years away.

Like fusion power, you mean?

No, exactly the opposite. This stuff works,

Only with human drivers available to take over.

and just needs fine-tuning.

There's general agreement in the industry that they "just don't work" in
snow. That's not "only-just don't work", flat "don't work".
--
Roland Perry