Slashdot Mirror
Self-Driving Cars Are Safer When They Talk To Each Other (engadget.com)
An anonymous reader quotes Engadget:
A University of Michigan public-private partnership called Mcity is testing V2V, or vehicle to vehicle communication, and has found that it makes their autonomous prototypes even safer. V2V works by wirelessly sharing data such as location, speed and direction. Using DSRC, or Dedicated Short Range Communication, V2V can send up to 10 messages per second. This communication allows cars to see beyond what is immediately in front of them -- sensing a red light around a blind curve, or automatically braking for a car that runs a stop sign... The catch of V2V? It has to be installed in the majority of cars and infrastructure (such as traffic lights) to function adequately.
If this was a Microsoft car, SMBv1 would be enabled by default.
Great because what we need is just to throw even more overcomplex, unpredictable tech at the problem rather than to simply get people to put their damn phones down when driving.
Some bad guy programs his car to give out false information causing lots of accidents. Would be a great way for a bank robber or other bad guy to slow pursuing authorities.
I drive a car with level 1 automation [speed control only].
Two weeks after I bought the car it paid for itself IMHO -- driving at dusk on two lane 50mph packed road; we were all doing 50mph (rare). I saw and was ready to take any action to a car (maybe two) pulling out making a right in front of me. I never saw the Jeep making a left into traffic behind the guy making a right. And then just didn't GO. ... "OH, now I see the Jeep". If I was driving I would have plowed into his ass end.
The car slammed on the brakes for me before I even saw the new car. I was more reacting to my car and what the hell is it doing
Two months later I was rear ended. Not bad; I do love that HEMI. :) The car's adaptive cruise control started the hard brake and the emergency braking system finished it off. The car won't stop itself 100%, but it will take you from 100mph to 10mph in short time / distance. Traffic hard stopped from 70mph in the left lane on a highway. I could see traffic stopping; let the car do it's thing better than I could. Otherwise I'd be further back giving myself for distance / time. The computer doesn't need it.
The guy behind me was way too far back and waiting far too long to HARD brake (more than I did IMHO). Unfortunately the car behind him wasn't ready and pushed his ass right into mine. My car stopped just short enough that after being pushed forward I was still 1' away from the car in front of me ... who at that moment pulled away as traffic was moving forward again. 1 second is all I needed.
Anyway -- wouldn't it be cool if my car could've communicated to the car that caused the accident (two back) and have its system start a nice slow brake to the stopping / stopped traffic. Re-adjust speed from 70mph to 35mph and maintain would've done it for the next group of cars...
It should be a trivial logical argument to suggest that "more accurate information leads to less error generated behaviour." One should note that the clause "accurate information" is a hell of stated requirement, but the measurement estimation of accurate true scores from error contaminated dimensions is a well studied field. Dealing with that in a online manner is more difficult, but a growing field.
The Spanish Inquisition of Psychometrics; Burning all the heretics.
I speak only to my car, and my car speaks only to God.
You are welcome on my lawn.
Good, now lets get beacons available to motorcyclists, so morons don't come in to their lanes, or turn left in front of them.
The Forbin Project comes to mind.
ELOI, ELOI, LAMA SABACHTHANI!?
And water is wet.
What, you think doubling processing power and number of sensors is not going to increase safety? Not to mention the fact that one of the cars is usually ahead of the other, so it gets information about static obstacles before it could normally see them?
excitingthingstodo.blogspot.com
A University of Michigan public-private partnership called Mcity is testing V2V, or vehicle to vehicle communication, and has found that it makes their autonomous prototypes even safer.
It will be safer until some asshat decides for fun or profit to screw with the system. It's easier to make a system safer when the hackers don't have access to it yet.
What worries me about a lot of this stuff isn't whether they can make the technology work but rather whether they can adequately secure the technology. I work in the auto industry and device security is simply something NOT a part of the engineering culture because it's never really needed to be. It's not that the engineers are dumb or are doing a terrible job but that they simply don't have a long history with making secure software. I think there are going to be some very expensive lessons learned the hard way before they get up to speed.
V2V is fine, but they better not put any significant stock in it - use like with gaming you cannot really trust what any external client sends without layers of verification.
Otherwise, way to easy to hack a fake obstacle the car has to stop for in order to hijack the car, or perhaps coordinate swerving of two oncoming cars so they hit...
"There is more worth loving than we have strength to love." - Brian Jay Stanley
George, so proud to receive one of the first truly self driving cars, sat down and spoke "Supermarket!" and the doors locked, off he went, and everything was fine until the car said, "Our records indicate your devices suggest you may be harboring an Unapproved Thought, please remain seated and do not attempt to leave the vehicle (recomputing) Detention and Reprogramming Center (recomputing)" and that's the last thing he ever remembered.
The catch of V2V? It has to be installed in the majority of cars and infrastructure (such as traffic lights) to function adequately.
What? No it doesn't, that's a blatant lie. No vehicle can trust what another vehicle tells it, that information can only be used for advisory purposes. Therefore, it only has to be installed in enough vehicles for a sprinkling of them to be following one another around in order to provide substantial benefits. And those vehicles are going to report on the state of traffic lights that they can see, so even some traffic light date will be in the system without any of them actually being explicitly connected to it.
In order to achieve the maximum benefits, yes, it has to be ubiquitous. And I expect that eventually, there will be laws requiring it — and by that, I mean before the human driver is outlawed on the public road.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Worried about sefl-driving cars not being able to predict or understand the actions of human drivers? Fear not, soon enough human-operated cars will go the way of analog TV transmission. You have a human driven car? Tough shit, buddy, your car is now illegal. You need to buy a Universal Enterprises self-driving car starting in 2021. No human-driven cars can be registered or driven after 1/1/21.
...when one party is malicious? Or lying?
If a bad actor lies to cars in a clever way, can he fool them into traffic jams, or accidents? What if a cleverly crafted purported message yields very high processing or requires more space than exists to decode, or even allows for injection of opcodes?
We live in a world where EVERY possible bad outcome for something being computerized and networked has been observed, and in many cases is being observed in the present. This is not some idle concern.
It is also pretty simple logic to know that communication between vehicles provides useful information for safe navigation. Dealing with the unexpected is a large challenge in autonomous driving, and such communication can reduce unexpected events. In order for it to work, there needs to be a standard and we should take time to do it right, along with other standards that would help autonomous travel such as intersection communications and probably even parking lot communications.
Self-driving cars who talk to each other don't need red lights for a safe crossing, that's one of the points.
So vehicles will talk to each other? Great, so this makes it easier to adjust cars behaviour based on other cars around it. So if, for example, I don't pay for the "gold" road plan my car can now safely pull over and allow the cars containing people paying the premium road rate (i.e. the rich minority) to power by unimpeded. Don't worry though, I'm sure I'l get shown plenty of adverts on my cars' infotainment system whilst I wait my turn.
Roads are currently a great equaliser; you and I use the same road as everyone else, sure cars might have different powers and abilities but your traffic jam is my traffic jam. Not for long though!
Standards like TAKE YOUR FUCKING EYES OFF YOUR FUCKING PHONE while entering a crosswalk, or just walking in general.
Good luck getting away now...
Comment removed based on user account deletion
Comment removed based on user account deletion
Comment removed based on user account deletion
If they start taking to each other then they going to start doing what traffic police do when they get bored - play snooker:
If your a pedestrian wearing red or black you should start worrying...
http://news.bbc.co.uk/2/hi/422...
Bad driving existed before and is caused by things other than cell phones.
I'm too lazy to compose a creative sig.
this should generate some interesting headlines lol
I have a new Turing test.
We'll know the long, slow caboose of our shiny new AI technology has fully arrived at the station when people no longer feel an irrepressible urge to posit that a lone bad actor can poison the entire system.
"N'uh — I don't think so," chime a thousand giant matrices in near-perfect unison.