Page 2 of 57

Re: Tesla's autopilot, on the road

Posted: Sat Jul 09, 2016 8:54 am
by edatoakrun
Two stories from AN lay out the significance of NTSB's autopilot investigation.

Odd, isn't it, that while many seem inherently fearful of self-driving cars before they experience them, the existential question once drivers experience autonomy, is how do you convince the same drivers that four-out-of-five-times-reliability is not good enough to turn over their lives to their autopilots...

NTSB to scrutinize driver automation with probe of Tesla crash

WASHINGTON -- For years, U.S. investigators have been calling for more automation on motor vehicles, such as sensors that slam on the brakes to prevent a crash.

At the same time, the National Transportation Safety Board, in its probes of transportation mishaps, has warned that such devices may also have a down side: the technology can confuse operators if it's poorly designed or lead to complacency that breeds its own hazards.

Now, for the first time in a highway accident, those two potentially contradictory themes will be put to the test as the NTSB opens an investigation into a fatal accident involving a Tesla Motors Inc. sedan that was driving with a feature called Autopilot enabled...

http://www.autonews.com/article/2016070 ... esla-crash

First real-world look at how drivers interact with autonomous technology

...NHTSA's inquiry is likely to focus on the ability of Autopilot's cameras and sensors to detect and react to road hazards. Even if the system performed as it was designed to, experts say, the agency could find that it poses an unreasonable safety risk as deployed.

Tesla rolled out its Autopilot system in October through software update, with key features still in the beta, or late testing, stage...

Tesla warned: "The driver is still responsible for, and ultimately in control of, the car."

The extent to which drivers heed that warning is unclear. Missy Cummings, an engineering professor and human-factors expert at Duke University, says humans tend to show "automation bias," a trust that automated systems can handle all situations when they work 80 percent of the time.

The Model S crash that killed driver Joshua Brown, a former Navy SEAL and Tesla fan, was "not an isolated incident," Cummings said, citing examples of other reported crashes involving Tesla cars that had Autopilot functions activated...

http://www.autonews.com/article/2016070 ... autonomous

Re: Tesla's autopilot, on the road

Posted: Sat Jul 09, 2016 11:18 am
by edatoakrun
They seem to have already made up their minds...

Consumer Watchdog Calls On Elon Musk To Disable Tesla’s Autopilot; Says Company Must Pledge To Be Liable For Self-Driving Failures If Feature Returns


7/8/2016

http://www.consumerwatchdog.org/newsrel ... pledge-be-

Mr. Elon Musk July 7, 2016
Chairman, Product Architect and CEO
Tesla Motors Inc.
3500 Deer Creek Rd.
Palo Alto, CA 94304

Dear Mr. Musk:

We are writing to express Consumer Watchdog’s concerns about Tesla Motors’ woefully
inadequate response to the tragic May 7 fatal crash in Florida of a Tesla Model S being
controlled by the autopilot feature, and the emerging pattern of blaming victims involved in the
crashes while using the feature. Our first concern is your inexplicable delay in announcing the
crash. You made a public acknowledgement on June 30, only after the National Highway
Traffic Safety Administration announced it was investigating. Such a delay when your
customers continued to drive on public highways relying on autopilot and believing it to be safe
is inexcusable...

Tesla is rushing self-driving technologies to the highways prematurely, however, as the crashes
demonstrate, autopilot isn’t safe and you should disable it immediately. If autopilot can
ultimately be shown to meet safety standards and is then redeployed, you must pledge to be
liable if anything goes wrong when the self-driving system is engaged.

Sincerely,

Jamie Court Carmen Balber John M. Simpson
President Executive Director Privacy Project Director

http://www.consumerwatchdog.org/resourc ... 070716.pdf

Re: Tesla's autopilot, on the road

Posted: Sat Jul 09, 2016 2:09 pm
by GRA
Not that they said anything that hasn't already been said by many of us in the other thread, but +1,000.

Re: Tesla's autopilot, on the road

Posted: Sat Jul 09, 2016 7:16 pm
by evnow
Similar features by other companies work even worse - and aren't in Beta.

Re: Tesla's autopilot, on the road

Posted: Sat Jul 09, 2016 8:04 pm
by DanCar
evnow wrote:Similar features by other companies work even worse - and aren't in Beta.
Does that include the hands off feature that allows people to doze off? Makes for a far more dangerous and useful system because of this.

Re: Tesla's autopilot, on the road

Posted: Sat Jul 09, 2016 11:09 pm
by jeffthewalker
[quote="edatoakrun"]Odd, isn't it, that while many seem inherently fearful of self-driving cars before they experience them, the existential question once drivers experience autonomy, is how do you convince the same drivers that four-out-of-five-times-reliability is not good enough to turn over their lives to their autopilots...[quote]

We (us humans) already turn over our lives to autopilots....

Traffic lights
Elevators
Fire alarm systems
Smoke alarms
An airliner with a pilot still has "automation" between the pilot and the physical parts

I can think of lots more. How about a list?

Re: Tesla's autopilot, on the road

Posted: Sat Jul 09, 2016 11:57 pm
by cwerdna
jeffthewalker wrote:
edatoakrun wrote:Odd, isn't it, that while many seem inherently fearful of self-driving cars before they experience them, the existential question once drivers experience autonomy, is how do you convince the same drivers that four-out-of-five-times-reliability is not good enough to turn over their lives to their autopilots...


We (us humans) already turn over our lives to autopilots....

Traffic lights
Elevators
Fire alarm systems
Smoke alarms
An airliner with a pilot still has "automation" between the pilot and the physical parts

I can think of lots more. How about a list?

The first four systems are orders of magnitude less complex than developing a level 4 or 5, or even level 3 autonomous vehicle.

(http://www.techrepublic.com/article/aut ... fferences/ and http://www.nhtsa.gov/About+NHTSA/Press+ ... evelopment explains the levels.)

A plane's autopilot also has FAR less to deal with given that planes are much further apart than cars, there aren't physical lane barriers, they don't need to worry about pedestrians, traffic lights, traffic signs, etc. There's also an air traffic control system, standards, along w/controllers and pilots that undergo lots of training besides a pilot's license being much harder to get than a driver's license.

The subject of vision in computers is a very difficult subject (most folks w/a computer science background know this) and there are all sorts of other challenges (e.g. interpreting traffic signals, recognizing pedestrians, bicyclists, gestures by police (and recognizing they're police vs/ a kook on the road) that might contradict traffic lights, etc.) In addition, you need to have reliable distance measurements.

See the IEEE videos at viewtopic.php?f=12&t=6338 or maybe the video in my 2nd post there.

As I posted elsewhere w/a few edits here:

There are so many questions and implications as cars become more autonomous, if not totally. Examples:
- who is liable in the event of an accident or traffic laws are broken?
- will drivers be able to react quickly enough and takeover in the event of the system screwing up or not knowing what to do?
- if there's a no-win situation where someone has to die, who should it kill? The driver or the others? What if it's between the autonomous vehicle having to send the car off a cliff to kill the driver and its passengers in order to save a busload of people?
- will drivers and others around them trust the technology enough?
- will driving skills atrophy so much that drivers will be unable to properly drive w/o aids or take over, in the event it's needed?
- how will autonomous vehicles inform other vehicles and pedestrians as to their intent? The Nissan IDS Concept video at ~2:25 https://www.youtube.com/watch?v=h-TLo86K7Ck shows examples of indicators and signage.
- should drivers be able to take over control at all or should there be no steering wheels or pedals or should they be disabled/retracted when in auto-pilot mode?

Re: Tesla's autopilot, on the road

Posted: Sun Jul 10, 2016 2:58 am
by LeftieBiker
Not really an argument for or against, but some of you may recall a lawsuit about a decade ago: an elderly couple bought an RV, and set out on a vacation trip. After getting on the freeway, they set the cruise control, and...both went into the back for a cup of coffee. The RV crashed, of course, and they sued, arguing that the salesman hadn't adequately explained how the cruise control system worked, and what its limitations were.

Re: Tesla's autopilot, on the road

Posted: Sun Jul 10, 2016 3:09 am
by jeffthewalker
cwerdna wrote:
jeffthewalker wrote:
edatoakrun wrote:Odd, isn't it, that while many seem inherently fearful of self-driving cars before they experience them, the existential question once drivers experience autonomy, is how do you convince the same drivers that four-out-of-five-times-reliability is not good enough to turn over their lives to their autopilots...


We (us humans) already turn over our lives to autopilots....

The first four systems are orders of magnitude less complex than developing a level 4 or 5, or even level 3 autonomous vehicle.


I agree 100% with every point you make.

I quoted "edatoakrun How do you convince the same drivers...." and gave a few simple examples where humans already accept many levels of automation.

I understand tech complexity.

Pilots license
Amateur radio license
Engineer (retired)
Microprocessor control programmer (still going)
Lay psychologist:-)

Re: Tesla's autopilot, on the road

Posted: Sun Jul 10, 2016 2:23 pm
by abasile
LeftieBiker wrote:Not really an argument for or against, but some of you may recall a lawsuit about a decade ago: an elderly couple bought an RV, and set out on a vacation trip. After getting on the freeway, they set the cruise control, and...both went into the back for a cup of coffee. The RV crashed, of course, and they sued, arguing that the salesman hadn't adequately explained how the cruise control system worked, and what its limitations were.

That one has sure made the rounds! http://www.snopes.com/autos/techno/cruise.asp