[SystemSafety] a public beta phase ???

[Smith, Brian E. (ARC-TH)]

Doesn¹t surprise me at all, Kevin.  I commute from NASA Ames to my home in
Mountain View and see unusual behaviors by the fleet of Google Lexus
vehicles and the little ³gumdrop² AVs equipped with just a joystick and no
steering wheel.  The Google safety drivers and systems engineers onboard
are fortunately not reading their newspapers.

I¹m on the Ames Safety Committee that is responsible for monitoring the
safety of driverless-car experiments being performed here by a major
automobile manufacturer.  The same subtleties and contradictions that have
been discussed at length on this thread play out in very practical ways
during our review of the activities here.

Brian

On 8/5/16, 1:50 PM, "systemsafety on behalf of Driscoll, Kevin R"
<systemsafety-bounces at lists.techfak.uni-bielefeld.de on behalf of
kevin.driscoll at honeywell.com> wrote:

>On Wednesday, I was passed by a Tesla on the 101 going to SFO from NASA
>Ames at about 4pm (rush hour).  The person behind the wheel (I couldn't
>call him the driver) had a stack of papers in both hands and was reading
>them.  He went by me too fast (40mph?, I was doing about 25) to see how
>often he looked up; I saw one head bob.
>
>> -----Original Message-----
>> From: systemsafety [mailto:systemsafety-bounces at lists.techfak.uni-
>> bielefeld.de] On Behalf Of Mike Ellims
>> Sent: Friday, July 22, 2016 06:54
>> To: 'Les Chambers'; 'Peter Bernard Ladkin';
>> systemsafety at lists.techfak.uni-bielefeld.de
>> Subject: Re: [SystemSafety] a public beta phase ???
>> 
>> Morning Les,
>> 
>> These are all on the surface reasonable ideas, however as Mencken said;
>> "for every complex problem there is an answer that is clear, simple,
>> and wrong"
>> (sorry - couldn't help myself ;-).
>> 
>> 1. The solutions suggested are applicable only to vehicles; what about
>> pedestrians, cyclists, horses, live stock, wild animals (deer or moose)
>> or any of the hundreds of objects that can be found on roads e.g.
>> roadwork's, fallen trees. To go beyond where Mercedes, BMW and Tesla
>> are now requires everything be detected.
>> 
>> 2. They don't scale, or at least don't scale fast enough.
>> First, to be truly useful all vehicles would need to be equipped.
>> Currently there are 1.2 billion vehicles on the worlds roads and
>> approximately 85 million new vehicles being added each year. Note this
>> doesn't include road going equipment such mobile plant or tractors.
>> 
>> Second the average life of a vehicle is somewhere between 10 and 20
>> years so even if we started mandating one of these solutions now it
>> would be 2026 or so before approximately half of the world vehicles
>> were equipped and 2040 before 100% (simple model assuming 3% of all
>> vehicles are removed each year).
>> 
>> Even if it was a simple retrofit like putting bumper stickers on the
>> front sides and rear of every vehicle it still need to be designed and
>> rolled out.
>> If it has to be done by a mechanic paid for by the manufactures it's
>> going to have a minimum cost of $50 US per vehicle... so we're talking
>> a minimum of $50 billion... And if it's bumper sticks they had better
>> be an exact match to the paint on my car!
>> 
>> But that would be unlikely to happen as,
>> 
>> 3. The solution would have to be legally mandated. This might be
>> possible Europe but in the US you would have to provide an economic
>> case that showed that the avoided cost in lives saved would exceed the
>> cost of adding the equipment to all vehicles. Three simple examples, in
>> Europe side impact bars on trucks, ABS and indicators a different
>> colour to brake lights are all required, in the US none of these are as
>> the economic case isn't clear enough for to allow the rule making
>> process to move forward.
>> 
>> For the Tesla that implies you need agreement between the EU,
>> USA/Canada, China and Australia (based on locations of the supercharger
>> networks).
>> Goodness know how long getting agreement would take but possibly
>> somewhere between 4 and 10 years?
>> 
>> I know this is a bit like pissing on the parade but the huge numbers,
>> time scales and politics involved make the problem approximate
>> intractable fairly well, which is why manufactures are following an
>> approach based on what they control.
>> 
>> It may be worthwhile to note that the vast majority of cars on the road
>> today don't need radar reflectors as they are radar reflectors, under
>> all that plastic they are steel (or aluminium in the case of Tesla). It
>> gets interesting with carbon fibre cars such as the BMW i3 but I assume
>> they have radar reflective material added to the mix as radar and
>> adaptive cruise control is so common today (you would hope that would
>> come out of the hazard analysis wouldn't you).
>> 
>> Remember in the Tesla crash the car's radar "saw" the truck (but only
>> the truck bed) but apparently misclassified what it was seeing... at
>> this juncture it appears to be a requirements or analysis problem not a
>> senor detection problem.
>> 
>> I suspect that fully autonomous cars are probably some way off, Google
>> is now saying 5 to 10 years, Musk says two but what he probably means
>> is 4 to 10.
>> 
>> Fully autonomous vehicles can probably be classed as a "wicked"
>> problem, in that the problem itself isn't well defined. Tesla and other
>> manufactures have been able to provide a partial solution for highly
>> constrained conditions, i.e. motorways and highways where the
>> complexity of the situation is relatively  "low". Noisy urban areas (as
>> an example) is a whole different problem. For example do we have to
>> detect people and classify them as people or is it enough to classify
>> them as things either moving or not moving; or perhaps as thing on
>> road, things moving towards road or things not on road? At this point
>> in time it may not be possible to answer that question.
>> 
>> 
>> Cheers.
>> 
>> 
>> 
>> 
>> -----Original Message-----
>> From: systemsafety
>> [mailto:systemsafety-bounces at lists.techfak.uni-bielefeld.de] On Behalf
>> Of Les Chambers
>> Sent: 22 July 2016 00:14
>> To: 'Peter Bernard Ladkin'; systemsafety at lists.techfak.uni-bielefeld.de
>> Subject: Re: [SystemSafety] a public beta phase ???
>> 
>> Peter
>> The concept of workshopping something is that you rollout ideas, people
>> critique them and in the process maybe you spark some creativity, new
>> ideas that is. These days our lives are absolutely ruled by the people
>> with ideas.
>> Elon Musk is a classic example (I'm in awe of that guy despite the fact
>> I disagree with some of the things he does). Ideas are the last
>> frontier, they are the final currency, they will never be automated.
>> In 24 hours My Tesla Motor Club post attracted 930 views and 22
>> replies. It looks like this is a very active forum biased more towards
>> solutions than the it'll-never-work-narrative - I find this refreshing.
>> The following three I found particularly informative:
>> ----response one ------------------
>> Good reference: Wireless Vehicular Networks for Car Collision Avoidance
>> http://www.springer.com/us/book/9781441995629
>> ----- response two -------------------
>> I'd favour a "radar reflector" of some sort - something that makes a
>> "vehicle" more "visible" to the sensors. Hopefully dirt-cheap, and thus
>> could be mandated for installation at the vehicle's next roadworth-test
>> (over here that is an annual test once a car reaches 3 year's old).
>> ------ response three ------------------ "high-quality GPS SPS
>> receivers provide better than 3.5 meter horizontal accuracy."
>> So i think accuracy is not good enough.
>> If there is a car on the side of the road, your GPS receivers are
>> closer each other than 3.5 m and collision warning would give false
>> alert.
>> ------------------------
>> 
>> Have you got any ideas Peter?
>> 
>> Les
>> 
>> 
>> -----Original Message-----
>> From: systemsafety
>> [mailto:systemsafety-bounces at lists.techfak.uni-bielefeld.de] On Behalf
>> Of Peter Bernard Ladkin
>> Sent: Thursday, July 21, 2016 3:51 PM
>> To: systemsafety at lists.techfak.uni-bielefeld.de
>> Subject: Re: [SystemSafety] a public beta phase ???
>> 
>> Les,
>> 
>> On 2016-07-21 01:44 , Les Chambers wrote:
>> > Ok, so I've posted my brilliant idea (below) on a Tesla Forum for the
>> > Model S. .... It got 4 views in the first 10 seconds after posting.
>> > Let's see how much interest it generates.
>> 
>> When I read your post I thought you were being facetious. But on the
>> odd chance you were being serious, a couple of comments.
>> 
>> First, fail-stop is a fairly well-understood mechanism, of limited use.
>> It is going to be of particularly limited use in road traffic, not only
>> because of its functional limitations but also because of the latency.
>> People's reaction time is between 1 and 2 seconds (this has been fairly
>> well measured with pilots). This is quite long enough to get you into
>> an irrecoverable situation in road traffic.
>> 
>> Second, the Tesla S is equipped with such a device. It's called
>> "steering wheel and brake activation" and it didn't save Mr. Brown.
>> 
>> Third, designing reliable GPS locator mechanisms, even for steadily-
>> moving objects, is tricky. For example,
>> http://www.icao.int/APAC/Meetings/2015%20ADSBSITF14/IP04_AUS%20AI.4%20-
>> %20Bo
>> eing%20787%20ADS-B%20deficiency.pdf
>> For general comments about the suitability of GPS-based devices for
>> high-resolution terrestrial use, see
>> http://www.raeng.org.uk/publications/reports/global-navigation-space-
>> systems
>> , the report of a Working Group chaired by Martyn. There is quite a bit
>> about trustworthiness and lack of it.
>> 
>> Fourth, car manufacturers have been working on such "sense and avoid"
>> mechanisms quite intensely for well over a decade. I recall a talk at
>> SAFECOMP 2004 in Potsdam from Daimler R&D guru Ralf Herrtwich on the
>> trustworthiness of automotive telematics. He was talking about car-to-
>> car stuff. It was mostly radar/lidar/sonar based, for what I take to be
>> obvious reasons, namely that you don't have to worry in quite the same
>> way about the trustworthiness of your sensorics as you do about the
>> trustworthiness of third-party information such as GPS positioning of
>> others. I doubt if that has changed at all.
>> 
>> I asked him about what they were doing about vulnerable road users and
>> didn't get the impression that they were doing much at all at that
>> point.
>> 
>> PBL
>> 
>> Prof. Peter Bernard Ladkin, Bielefeld, Germany MoreInCommon Je suis
>> Charlie
>> Tel+msg +49 (0)521 880 7319  www.rvs-bi.de
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> _______________________________________________
>> The System Safety Mailing List
>> systemsafety at TechFak.Uni-Bielefeld.DE
>> 
>> 
>> ---
>> This email has been checked for viruses by Avast antivirus software.
>> https://www.avast.com/antivirus
>> 
>> _______________________________________________
>> The System Safety Mailing List
>> systemsafety at TechFak.Uni-Bielefeld.DE
>_______________________________________________
>The System Safety Mailing List
>systemsafety at TechFak.Uni-Bielefeld.DE