Potholes abound on the road to car-to-car communication | Hacker News Reader

Potholes abound on the road to car-to-car communication

arstechnica.com

13 pointsleoedin21 comments

Not mentioned in the article, but perhaps a relevent comment on the privacy concerns: In the UK at least, the government have already installed license-plate tracking cameras on overpasses on almost every major road in the country. Sniffing the RF V2V data wouldn't really reveal much more (apart from speed).

I suspect that it would be politically unacceptable to routinely prosecute for speeding offences that were identified with this data (although perhaps there would be a grey area when very high speeds were identified).

It could transmit everything collected by your car's event data recorder. This includes information about your car maintenance, current seatbelt usage, derivable information about how much weight your car is carrying, etc. It basically creates a data channel between any onboard computer in your car to a wireless network.

Some cities already do this (Seattle for one) with red light cameras, license plate readers, and automated fines.

>I suspect that it would be politically unacceptable to routinely prosecute for speeding offences that were identified with this data

Authorities issue speeding citations to fast drivers all the time based on toll road time-stamps.

You shouldn't be surprised about the government's actions when there's a windfall at stake.

No mention of the impact to bicycles and pedestrians if people start to rely on these systems. Installing these on bikes might help a lot, making them more equal.

It would bring surface transportation up to par with sea (AIS) and air (ADS-B) transport.

In the docs I read a few years ago, there was talk about integrating the system into handheld devices like cellphones and portable GPS gizmos. It would definitely be possible.

My first thoughts after reading the NTSB press release:

* Police are going to love this. Why sit in your cruiser and fire a radar/laser gun (which requires aim, patience, concentration , and a is error-prone) when you can just sit back and let the cars themselves tell you when they're speeding. For the same reason, once people figure it out, consumers are going to hate this.

* Hacks will abound - some good, some bad (you get to apply the labels you want ;) For example, I can imagine a car being hacked to report 99mph when going through intersections so everyone else stays out of the way, but if a police officer is nearby (thank you, Waze), 55 allllll the way home.

"But what worries some is that the system could be used to target “bad actors” on the highways, as they were referred to in a recent Government Accountability Office (GAO) report. Cars may become informants on drivers who speed or drive erratically"

This is a feature, not a bug. Cars should become informants on dangerous drivers.

        INT.  CAB - NIGHT

        Korben tries to control his car, reeling from the impact.

						VOICE (O.S.)
				You have just had an accident.
				Seven points have been temporarily
				removed...

        Korben manages to stop his cab, pulls over to the side.

						VOICE / KORBEN
				You have one point left on your license.
				Have a good day.

I'm working on a similar system. The impact this will have on the tech industry is potentially huge. Imagine your mobile device pairing with cars. All of your preferences are automatically uploaded to the vehicle. Including predetermined destinations that you might have already defined before leaving. As you move down the road, the applications you have installed on your mobile device will give you data regarding the surroundings. Yes, it will probably contain advertisements, but it will also contain valuable information about the area.

I know that privacy is a concern (which is shared among all). It is the biggest challenge faced at the moment. Implementation is heavily dependent on third party systems over which we don't have any controls.

V2V is a great idea, maybe the equal of airbags, or ABS. But if I can't actually trust the government to implement it, well...

In this case I trust the government to implement it more than I trust the world to never create a tiny, throwable "I'm going straight in your direction at 200mph" beacons. Unfortunately that's just a question of time.

Encrypt transmissions with NTSB provided secret keys.

In that case, the point of least resistance moves to spoofing the accelerometer readings...

Yeah... encrypt the transmission with a secret key that is embedded in every public car. That will stop it being reused! Like CSS, bluray and other DRM solutions.

There is a argument against such help, because the drivers will learn to rely solely on the warnings and stop properly monitoring the road. As soon as the system fails, accidents will start to happen.

Accidents already happen, and frequently. The question is "will these systems fail enough such accidents happen more often than they do now?"

That is different from my experience. If an automatic system that replaces a human is adopted, it has to be almost perfect in the eyes of the public and regulators. For example; plane autopilot is more reliable that the pilot, yet we still have them( pilots ).

Yes, and autopilots are absolutely terrible at dealing with unexpected events. See: UA232, BA38, US1549, AC143. In each of those accidents, passengers survived who would have died in the hands of an autopilot.

1) In the BA case, the autopilot blindly commanded full thrust and then pitched up in spite of decreasing airspeed. Humans correctly determined that the engines were not producing thrust and elected to make an off-runway landing instead of stalling the plane into airport-adjacent houses.

2) In the UA case, an engine failure disabled all three hydraulic systems, leaving the plane without primary flight controls. The autopilot had no provision to command differential thrust from the two working engines, and would have driven the plane into the ground.

3) In the AC case, humans flew the plane at best L/D after running out of fuel, which enabled the airplane to land at a closed airport (any airport database would have told the autopilot it couldn't land there).

4) In the US case, humans determined that making an airport landing would be impossible and elected to land the airplane on the Hudson river, which has (and never will have) a published approach procedure.

Please explain how you expect autopilot programmers to anticipate these scenarios and correctly chose a solution.

We can't. Thank you for making my original point, which was that human drivers are very hard to replace.

But wouldn't this turn every vehicle into a moving GPS beacon for drone strikes?

Reply on news.ycombinator.com