NHTSA to require vehicle to vehicle communications

[strider]

May not let you tailgate

All you need is one doofus in an '83 Camaro or slammed '92 Civic to screw it up for everyone. I've always thought the only way this autopilot stuff will work would be on separate lanes where cars would have to have the proper equipment to enter.

 

[brianman]

All you need is one doofus in an '83 Camaro or slammed '92 Civic to screw it up for everyone. I've always thought the only way this autopilot stuff will work would be on separate lanes where cars would have to have the proper equipment to enter.

So you have several AOV lanes and a single, undermaintained LOV lane.
AOV = Autopilot-Only Vehicles
LOV = Lame Old Vehicles

 

[RubberToe]

I'm sure this will be part of the implementation, but just to throw this out there...

Typically electronics in cars is several generations behind mobile devices, and generally not upgradeable. Look at the Entune system in a 3 year old Toyota Prius versus what your 1 year old cell phone can do. My understanding is there is a RF spectrum allocation that has been set aside for the wireless portion of the system. Both the hardware and the firmware/software on these systems needs to be easily upgradeable. Otherwise, you get into a situation where you might obsolete a whole bunch of cars that were built with Version 1.0 and version 2.0 is so much more advanced that Version 1.0 is no longer usable if it isn't upgradeable. Kind of like how USB went from 1.0 to 2.0 to 3.0, which still being backward compatible, though the newer versions still talked to legacy hardware.

I would imagine that part of the system definition will be that the software will be (must be?) able to be upgraded over the wireless network. If they discover some catastrophic flaw in the code that ends up killing someone, you want everyones car software upgraded immediately if not sooner. I would also imagine that there will be a set of defined hardware sensors that will collect data (speed, direction, position, etc.) that will be broadcast to other vehicles. As sensor technology improves, they might add newer sensors to the platform (slip detection for ice/rain, video stream, etc.). If the data for these must also be communicated between vehicles, the system will need to allow for this too, while still communicating with vehicles without the upgraded sensors.

I saw an interrview with a guy on Newshour last night I believe where this was being discussed. I concur with his assessment that there is really two different ways to approach the eventual transition from manual driving to fully automated vehicles. One involves what is being discussed here, where vehicles communicate with each other, which could probably eliminate maybe 50-80% of human error accidents. The other approach is to put all the intelligence in each vehicle, kind of like it is now where the "intelligence" is the human doing the driving. Though the two approaches aren't mutually exclusive. A completely smart car could certainly make use of other vehicles data, where an accident that just happened 2 miles ahead on the freeway would alert the trailing cars to get off the freeway ASAP.

Interesting times we are living in, thats for sure.

RT

 

[ElSupreme]

Well first of all, they can make modifications of the system illegal. Similar to how the telecommunications act made esn cloning illegal for cellphones. Second of all, they can still pin you with reckless endangerment, obstruction of traffic and etc.

Obviously 'reckless endangerment' might not be protected speech ('fire' in a theater). But obstruction of traffic, I think not. I didn't obstruct anything if I say things.

Spoofing wirelss devices is incredibly simple. You could make a phantom car or multiples that don't really exist.

And this would be covered by the first amendment and 'free speech'. And this was more what I was alluding to. But also telling people I am going slower than I am really going isn't dangerous, and even bold face lies are protected speech.

When the car realizes that it was inaccurate, you can have the car take a picture and identify the license plate of the car not matching the signal. So again they would get caught. Now multiple cars might be a bit tricky because how would they spoof the direction of the signal? Especially if they put antennas on each side of the car and make them directional. And say hypothetically they do, when your other sensors notice that they would report it.

On top of that, if each car has the unique identifier built into the hardware of the car. And the signal is sent encrypted, then decrypted at the cloud, you would have no way of spoofing it. Even if you somehow did spoof it, there would be authentication and the system would realize that a car exists in 2 places at once. Which again would trigger a flag.

At end of the day, the effort would not be really worth it.

Going to a 'cloud' with this sort of thing won't happen. You are talking too much of a lag for real time data use. A half second and I am 50 feet away from where I was previously. It is going to have be car to car. Not to mention the privacy concerns about logging this sort of data, with unique identifiers.

 

[stopcrazypp]

It think the idea is this is an early warning system and you/your car has the discretion on how to respond. If you drive a regular car with this system, it gives you a warning (and you can judge if it's fake or not).

If you drive an automated car, then it'll respond automatically, but the tailgating thing might not even be an issue in that case.

I think giving out fake signals will be similarly prosecuted as messing with traffic signals (similar effect).

 

[ggr]

Spoofing wirelss devices is incredibly simple. You could make a phantom car or multiples that don't really exist.

I went to a talk at the Real World Cryptography conference a few weeks ago, and one of the good presentations was about the crypto protocols for this standard.

. The car's identity is somewhat protected, in that it can create different temporary identities quite often.
. Each message is digitally signed in a way that can be verified by all the other cars
. ... but having said that, IIRC, emergency messages like "brake NOW" might be actioned before they are verified
. the signatures from the temporary IDs can be traced back to the real IDs by appropriate authorities.

(The talk was by William Whyte at RWC 2014. You can find him if you try. The standard is not yet fully baked, so I don't think his slides are available yet.)

Having said all that, I think this is a bad idea. The one demo we saw at my work had the car stopping in the middle of car parks because some parked (but turned on) car was pointing in the wrong direction. If you run the numbers about bandwidth, reliability of transmissions, and how many cars might be trying to talk at the same time on a crowded freeway, it's pretty easy to conclude that it will work just long enough to be mandated, and then stop working. And a single jammer (yes, illegal, but how will you find the guy in the middle of the multi-car pileup?) could cause great havoc.

 

[Denarius]

The problem with cars is that people keep them to long. If they build a car now, the processing power we will have in 10 years will probably be able to brute force the key in a short period of time. Not to mention hacking the vehicles implementations directly. High tech companies have a hard enough time securing devices, look at mobile phones. Now tell me that Ford is going to do a good job.

It can be done right, and it probably won't be much of a problem, but in a peer to peer wireless scenario I'd have no worries at all of getting caught tinkering with the system. Cars won't be triangulating signal, they will likely pass GPS coordinates in their signal.

I will have to look up William Whyte's talk.

 

[GSP]

?......But also telling people I am going slower than I am really going isn't dangerous, and even bold face lies are protected speech..

This may not be dangerous today, but consider the effect of your car's transponder broadcasting false velocity information. In the future, other cars may decide it is safe to cross a highway, but because your car is actually going faster than what other cars think, you will T-bone them. This could result in serious injury or loss of life.

In order to enable autonomous driving in the future, transponders will have to be mandated in all new cars long before there are any autonomous cars to listen to them.

GSP

 

[AudubonB]

And the 130 million (guessing) vehicles currently on US roads do what?

This seems to be something that might be a great idea for a system starting de novo; how to deal with that incredible mass of extant vehicles, though, renders it truly impractical in my eyes.

 

[physicsfita]

It think the idea is this is an early warning system and you/your car has the discretion on how to respond. If you drive a regular car with this system, it gives you a warning (and you can judge if it's fake or not).

That's exactly what the test car that I drove in the research study did. It had special blind-spot detection warning lights. If a transponder-equipped car was passing me, a light next to the appropriate side mirror would blink. I loved it, and I miss that feature. The collision-alert system that was installed was in an early state, and had problems with false alerts. It got better during with each update during my involvement with the study. At all times, and braking/steering decisions were mine alone -- the alerts were strictly informational.