Slashdot Mirror
U. Michigan Opens a Test City For Driverless Cars
An anonymous reader writes: The University of Michigan has opened Mcity, the world's first controlled environment specifically designed to test the potential of connected and automated vehicle technologies that will lead the way to mass-market driverless cars. Mcity is a 32-acre simulated urban and suburban environment that includes a network of roads with intersections, traffic signs and signals, streetlights, building facades, sidewalks and construction obstacles. The types of technologies that will be tested at the facility include connected technologies – vehicles talking to other vehicles or to the infrastructure, commonly known as V2V or V2I – and various levels of automation all the way up to fully autonomous, or driverless vehicles.
It's about time driverless vehicle testing has moved away from the snow-less climes and into an area that presents some actual challenges to the driverless vehicles.
Gary, Indiana, would have been perfect.
What happens when the city power goes out and all of the stop lights go out -- not blinking red, but off? I'm guessing the cars know where stop lights are supposed to be. Maybe they just pull over and wait. Is it smarter than a 5th grader? What about when 4 cars get to a 4-way stop intersection at the same time? Then what do they do?
....go wrong with v2v communication?
All it would take it 1 compromised vehicle to gain access and cause complete and utter chaos.
Humans, for all their faults, are for the moment still not remotely hackable. Until this changes, I do not see automated vehicles becoming the norm any time in the foreseeable future.
Unfortunately I think it will take a major disaster before the majority of people realize this.
Just use Detroit: it's full of real roads and building, full of perils, and many parts of the city are virtually devoid of people.
"A door is what a dog is perpetually on the wrong side of" - Ogden Nash
Testing autonomous technologies in public areas is essentially irresponsible. It goes from Google, who is extremely professional, but still gets rear ended a lot because the vehicle is over cautious and stops where humans would not to Delphi on down where I don't trust a damn thing they do and they really shouldn't be allowed in public at this time.
It's hard enough doing simple things like straight line freeway driving that it's seen as a major accomplishment when no one dies and they can out do drunk and distracted drivers under all conditions but only under clear sunny free flowing freeway driving. These courses, akin to closed courses for learning human drivers, are a necessary first step in bringing this technology to practical use.
The real problem with autonomous driving is environment variability, like road problems, debris, construction, animals, children and pets, other human drivers etc. when a computer 'sees' a cyclist they may or may even not recognize its a cyclist (ie maybe it assumes pedestrian given its sensor history) and form some standard metrics about the expected behavior. When a human does you immediately see if you make eye contact (so you know they see you something no vehicle does to my knowledge today), if they are on a pricy bike and clothed in biking gear or in dirty clothes in a beater and looking drunk - this allows humans to more accurately predict the future using things computers can't do for at least a decade out. For all the "nanosecond" decisions in urban settings computers can do humans are light years ahead in path planning and pattern recognition. Anyone in the field gets an immediate appreciation of how their toddler far exceeds a supercomputer and 500k in sensors even in 2015.
DARPA opens test city for weapons and advanced armor technologies. https://en.wikipedia.org/wiki/...
Pretending this is my office full of bitter coworkers..
This, just two hours after the headline of "Remote Exploit On a Production Chrysler To Be Presented At BlackHat"
SEEMS LEGIT!
Seriously though, I'm all for automation, communication between devices, and anything else that can help optimize any system out there (this being the transportation system as a whole, not just the internal computer system within a single vehicle) - BUT we need some serious security and accountability. We've seen serious issues with cell phone manufacturers delaying security updates by months (sometimes YEARS) to handsets. Is this going to be the case with cars, too, which traditionally never even get software updates to their infotainment systems? Maybe now that these systems can impose serious risk of physical harm to individuals both inside and outside the cabin, can we get some mandated government regulation that forces software security updates on a regular basis, that also persist beyond just the first year or two of the vehicle?
TLDR: Let's get this shit security, and continue securing it over time with updates, and I'm all for it!
Now that's driving.
First time they deploy these in public, some small kids will get killed and it's game over for the auto companies. Parents with dead kids don't ever forget.
Ever.
-- Tigger warning: This post may contain tiggers! --
Seriously, why should automated cars be operating in places where kids on bikes and skateboards and pets exist?
Automated cars might make sense in secure retirement communities, where everyone is 55 and nobody minds if they die.
-- Tigger warning: This post may contain tiggers! --
Get in touch when you turn 55. Let's see how ready to die you are.
You are welcome on my lawn.
If you want a real world test, it has to be where kids, with or without bikes and boards, and pets dart in and out of traffic. Some of the 55 y/os might be dumb enough to do that, but they're too slow to make it a proper test.
Slow down, cowboy! It has been 4 hours since you last posted. You must wait another few hours.
THIS is why we need dash cams. Good ones, looking fore and aft with a wide enough view to deal with stuff like the cyclists and biker's lane splitting shenanigans.
Well, it's something that tends to be integrated into self-driving cars. Google has very good views of all the accidents it's cars have been in.
Since I don't reply to AC's:
1. Can't do much about cyclists other than to build your mirror a little more sturdy. On the other hand. Self driving car! It wouldn't have a mirror there, because it uses a camera & other systems, and isn't stuck in the 'driver' seat.
2. Given self-driving car reflexes, is going to result in them eating a lot of pavement for no result. Or they cut it so close that the car still runs them over, causing serious injury, but the company insuring the car also makes the drive system, and points out they have video of them diving into traffic.
3. Somebody following you? Call the police with your cell phone or use the vehicle communication button(on-star for some brands).
4. They claim, you show the black box footage, they end up paying, and maybe spending some time in jail for lying before the court.
5. Seriously? Do you know how ridiculously dangerous getting into a 'deliberate' accident on a motorcycle is, even if you're wearing full gear? Again, self-driving car, it avoids the accident, the moocher has to find his insurance lotto elsewhere. If he does manage to make you hit him, black box will show it and he gets to pay his own medical bills.
6. Toll booths - EZPass anybody?
I don't read AC A human right
Seriously, why should automated cars be operating in places where kids on bikes and skateboards and pets exist?
Because reaction times measured in the milliseconds are handy at avoiding hitting them when they dart into the road, much better than people with reactions upwards of a full second?
I don't read AC A human right
Why are we testing on humans?
Now, hamsters ...
-- Tigger warning: This post may contain tiggers! --
They have a real football coach to test all this with, too?
Autonomous cars arent going to be better than humans at everything, but in the long run they will have far fewer crashes because they pay attention 100% of the time and have fast reaction times.
Humans are terribly unreliable. Yes they can foresee some certain types of incidents, but only if they are paying attention, aren't using their phone, driving drunk, putting on make-up, screaming at their own kids in the back seat, eating, of about a dozen other things that make humans terrible drivers regularly.
The grandparent was challenging the specific claim YOU made that SDC's get rear-ended because they stop too often.
The article you linked doesn't even mention any of the collisions the SDC's have been involved in, and certainly don't support the claim you made about them.
Try again.
Ok mr coward here are the actual citations, I was quick to post and put the wrong post above. Too many hipsters with common place movie knowledge and not actual engineerinrg knowledge so here you go I hope someone actually reads this that will appreciate facts.
The number of miles that google cars have driven so far including its safety record as stated by google directly : http://venturebeat.com/2015/06...
This equates to an accident every 90.9 thousand miles though https://en.wikipedia.org/wiki/... the wiki clams 14 accidents that puts this at 71.4 thousand - all of which are meticulously planed courses - none have left turns into traffic - none of which are in heavy traffic at all - none of which are in adverse conditions such as snow or rain which the car can't even function at all and which of course makes for more difficult human driving. So it is pointless to try and compare that to bumper to bumper rush hour driving on freeways, hundreds to thousands of accidents during large snowstorms, bad winter driving in general, rush hour in general, etc. This makes it extremely difficult to compare to actual human driving statistics.
Note the wiki even states google as saying the car will often revert to extra cautious safety conditions and cannot handle many situations which wont even be addressed till 2020.
Google admits its cars rear ended suprisingly often : http://www.dailymail.co.uk/sci...
Given that google has not made more than one accident detail public this is hurting their image in all likelyhood. Its a pretty safe assumption that a slow plodding car that often stops, say like you are supposed to text book style, but no one does, for pedestrians, will cause motorists to rear end you.
that it is in Rick (not Dick) Snyder's neighborhood.
Listen dumbass who wont use their account i said vehicle is over cautious and stops where humans would not which is exactly what the wiki says. Humans dont follow textbook rules the car does. Post using your id at least coward. Also you are a dumbass.
The Wikipedia article doesn't say that, though. It does say that, in 2014 at least, the Google cars would become extra cautious (not 'over [sic] cautious' as you claim) should they approach an unknown, complex intersection.
Then Google can teach the cars to do just that. You seem to think there are things these cars can't even conceive of, which is patently nonsense as they are trained by humans. If there is a great way to behave in a certain situation, the car can be trained to do just that, but with its far superior reaction times.
Try your test with human drivers and see how many toddler dummies you get through.