Brad Ideas Robocars &amp; More

Sat, 2019-11-23 07:33

Infrastructure

C-V2X, or something like it, seems like it could have some uses. I think the state governments would need to get on board first. Roll it out to the emergency vehicles and the construction vehicles and the various temporary traffic control devices (maybe not every traffic cone but at least to the large arrow boards). Roll it out to the traffic light systems (it's still not infrequent to have my view of traffic lights blocked by trucks). Eventually cover the entire road network. Allow routing, not just direct communication, but only by packets signed with the proper keys. (Maybe all participants, even mobile ones, can be required to route certain types of data packets?)

Once at least a few large states really adopt it, I think the private industry will voluntarily adopt it.

I think some portion of the spectrum would have to be licensed in order for that to work. I think you need to let the states control it, and specifically that they can give priority to certain users (emergency vehicles, construction vehicles, traffic control devices, ...) as well as mandate the proper protocols.

Unlicensed is probably fine for vehicle to vehicle. I don't see much use for it today, but in a robocar future it'll be useful for communication between robocars. A lot of that will probably also go over the Internet, but you don't always have Internet access, plus you could use direct v2v to communicate your IPv6 address. (I should say to communicate one of your IPv6 addresses, as you'll probably want to rotate them frequently.)

"I'll pay you 0.00001 BTC if you let me pull out of my driveway in front of you" would be awesome, and now that I think of it, a lot of it would work with an ADAS car.

Sat, 2019-11-23 10:56

Not low latency

Most of the applications you describe do not need extremely low latency, <20ms. Applications willing to accept longer latency with more reliability are better done through existing data networks, particularly traffic light timing and control.

Sun, 2019-11-24 04:54

Hmm

You must have inferred the applications from my partial list of devices, because I didn't state any applications except the Bitcoin one (which could be done through wifi direct, I guess).

Which "networks" would be better for state-controlled communications between vehicles, traffic control devices, traffic lights, etc? That's not rhetorical. I am probably missing something.

Low latency isn't the only advantage to short range direct communication. Privacy and better protection against DOS attacks are two others.

You wouldn't want to hook a traffic light up to the Internet to allow everyone in the world access to request a light change. But now I'm inferring what you meant by "existing data networks."

As I said in my original comment, most of the communication could take place through the Internet. But you often do need some short range direct communication to exchange IP addresses and/or public keys, if you want to maintain privacy and protection against DOS. Maybe just a tiny sliver of bandwidth is needed for that?

Maybe there's already a network out there than can authenticate location and route based on location. I'll do a Google search for it.

Mon, 2019-11-25 00:55

What networks for communications between lights and control etc?

Whatever network is out there and works. Most people use the internet for their communications today, not because anybody mandates it, but because it's the right choice for the job. The internet, of course, is not a physical network, but a collection of them and protocols to glue them. The law should never mandate hardware or protocol decisions if we can avoid it.

Direct communications are terrible for privacy, because you end up having to do public broadcast of your information to anybody, rather than encrypted broadcast to people you trust, who only release it to those who need to know it. The DSRC BSM has every car constantly transmitting where it's going to everybody. They have it roll the ID every so often (ha ha) as though that's going to solve the privacy issues. You are the first I have heard to claim it's better for privacy -- everybody else worries it's worse.

Protection against DOS? DOS is trivial against DSRC, though you do need to transmit in the area -- or take over a transmitter in the area.

I would not hook up a traffic light to "the internet." I would have a traffic light talk only to its trusted master over a secured and authenticated channel (which can run over anything, including the internet.)

There are some systems that do authenticate location of the transmitter of DSRC signals. That's a useful thing (if you are going to use these signals, which is not very likely.) But it's probably better done by things like local cell towers, that can do it better.

yes, there is an appeal to doing it yourself and not needing to trust anybody else, but it has so many overwhelming downsides that it's not enough to save the day.

Mon, 2019-11-25 06:54

Whatever network is out there

Whatever network is out there and works? Is there one? I guess you could use wifi direct (which isn't a "network") for the P2P stuff, but that's kind of convoluted, and I'm not sure it'd scale to the number of devices you'd need it to scale to. And what about for communications with traffic control devices and the like that should use a licensed band to protect against interference?

Your other criticisms seem to be of DSRC. My comments never even mentioned DSRC. I referred to "short range direct communication" as a generic term, not as a specific protocol.

There's no need to broadcast "your information" to "everybody." You only have to broadcast the information you wish to communicate to the others within range of the signal. The signal could be, and maybe should be, directional, which would further cut down on the possibility of easedropping or man-in-middle attacks. But unless you want to introduce a trusted third party or want to print public key QR codes on bumpers or something, there's no way to avoid that completely. (Maybe we can flash the high beams in Morse code? Only joking partly.)

You can't do "encrypted broadcast to people you trust" when communicating with strangers, unless you are going to have a trusted third party facilitating all communications (hence destroying privacy). The whole problem is that you need to be able to communicate with random parties, who happen to be in range, that you don't trust.

We need a way to allow "road devices" (vehicles, traffic control devices, etc.) to communicate with each other that is similar to the ways humans driving cars communicate with each other. Perhaps the radio spectrum isn't the best way to do that, but I'm not sure of a better plan.

Short range, direct communication facilitates privacy. It also helps protect against DOS. As you note, you have to be in range to perform the DOS. It's not possible to completely eliminate DOS when using radio communications, but you can eliminate it for practical purposes by requiring the attacker to be local (or, at least, to control a local transmitting device).

Maybe we should take an example. How should a car tell a traffic light, "hey traffic light, I'm approaching you, and I plan to make a left turn"? I think it should be direct, short range communication, not routed through a network or even through a cell tower.

Same question for "hey red car in the right lane, please yield to me."

Mon, 2019-11-25 10:56

How to

I have a bunch of articles on this topic which cover many of these subjects.

As for "hey light, I am approaching" the best way to do that is have a camera on the light that sees your left turn signal.

I am not sure what the other example means, other than a turn signal again.

Generally, remember, I don't think cars should communicate with strangers at all. While it has minor benefits, it is really hard to make work and presents security and privacy risks that don't justify those benefits.

Mon, 2019-11-25 15:14

Communicating with strangers

It only has security and privacy risks if you're doing it wrong. The drivers of cars communicate with each other quite frequently. They do it not through a network (which is what would introduce the security and privacy risks) but through short range, direct communication (sometimes sound, but usually using the visual spectrum).

Using radio communications, self-driving or ADAS vehicles could communicate even more information even more accurately.

I guess you don't see the benefit of it. Maybe because your view of the future or driving is a small number of oligopoly robotaxi companies. I suppose if that's the reality, peer-to-peer communication between strangers becomes unnecessary. But I don't know that that'll ever be the reality, and it definitely won't be the reality for a long time.

Mon, 2019-11-25 16:00

There is some benefit

But it's modest. And that does not outweigh the massive problems with getting it to work and the downsides. I have challenged all V2V booters to tell me what these big benefits are -- that are not attained by the already working and easily adaptable Vehicle to cloud to vehicle approach -- and so far they can't show much. Yes, V2C2V doesn't work out in deserted rural places, but car cooperation isn't all that valuable out there.

Mon, 2019-11-25 18:01

What about privacy?

Does the V2C2V system that already works involve a trusted third party in the cloud that is able to track all the vehicles?

Mon, 2019-11-25 19:05

How trusted

Well, the several existing ones (Waze, Google Maps etc.) certainly do, but it is possible to build systems that don't, and a good idea. Systems which don't retain information, or retain it only on the user device. Systems which only send you the information you actually need rather than broadcast everything about everybody.

Or it is possible to have "cloud" mean "local cell tower" and have it just relay the information that used to be broadcast, forgetting it quickly, with the advantage that you can have reliable transmission. (The tower can know if a message was lost and get it retransmitted, and it can coordinate transmission so individual transmitters never step on top of others, among other things.)

But while the cloud approaches have their flaws (including privacy risk) the fact that they actually exist and work gives quite a leg up over things that don't exist and are not yet shown to work.

Tue, 2019-11-26 06:01

Cloud vs p2p

That does seem to be the general answer. Yeah, cloud-based solutions are a privacy nightmare. But they're easier than doing things the right way.

Hopefully that'll change.

Tue, 2019-11-26 10:04

Not just easier

They make easy what is next to impossible. Tell me of a networking technology that was a success that did not provide value to the early customers? Fax machines sold because one company could buy 2 and put them in two offices and fax -- only later did you get to fax other companies. Same for phone lines, and everything else. Waze started by giving you navigation, later you got the effect of information from other cars. V2V provides almost no value until tens of millions have it.

Anonymous

Sat, 2019-11-23 13:47

Proper SI prefixes for mega-

Proper SI prefixes for mega- and giga-hertz should be MHz and GHz. m is SI prefix for mili-.

Sat, 2019-11-23 16:03