I doubt there will be an issue with this. Regulations will require the communication. A building is only a temporary delay in the transfer of information. Communication between self driving cars is the key to their greatest advantage in reducing traffic problems.
OTOH we’re getting well ahead of ourselves in how fast these things will actually be on the road. There will be plenty of incidents just in the testing phase that will require more development and spur the creation of new laws and regulations that have to propagate through the 50 states.
Well, one theoretical advantage of self-driving cars is that they should be able to communicate between themselves (at a bandwidth considerably greater than the honk or raised digit that humans are limited to). So if one car encounters a pothole, it should send a message to neighboring cars, conveying the digital equivalent of “Dammit@! Look out for that pothole on Rt. 5 near the Esso station.” One could imagine a database of road hazards – one that would be hopefully updated at intervals, so you don’t have cars swerving around imaginary potholes twenty years after the pothole has been filled.
Are there any serious proposals out there yet to create a centralized network, and to require that all SD car companies participate in such a network once it is created?
But what about the piece of plywood with nails it, that has shifted 15 feet down the road after being plowed over by cars unwilling to avoid it, if they even see it?
You heard it here first folks; we will never have fully autonomous cars on our current infrastructure.
We may have a carefully maintained parallel interstate system, using some sort of beacon system to keep the car on track, but there’s no way there will be a comprehensive network of every road made autonomous-capable.
While it’s probably not optimal, I see no reason why automatic pothole detection couldn’t be based on a regular camera. I mean you don’t detect potholes with LIDAR, do you?
I don’t understand that in another thread about skipping commercials, the consensus was that you’d need perfect AI in order to get a computer to recognize commercials, but somehow all the extraneous things that go on while driving will be implemented in the next 5 years? I just don’t see it happening.
What TriPolar said. There will be a federal standards and enforcement office. Whether it will fall under DOT or FCC or something totally new remains to be seen.
Not that I’m aware of, but there will be. All the car companies will be sending an engineer or twelve to mold the requirements to benefit whatever tech they’ve patented.
I instantly read this as “Straight Dope car companies”.
Anyone who thinks autonomous vehicles will never see widespread public use is a stoneheaded pessimist. The problems are solvable, and the benefits (energy savings, safety, and roadway carrying capacity) are humungous.
OTOH, anyone who thinks autonomous vehicles will see widespread use within the next few years is bizarrely optimistic. Some of the problems have been solved or nearly solved, but some of the remaining ones are difficult - and the transition from all manually operated vehicles to all/mostly autonomous vehicles is very daunting.
Consider a SD car being mass-produced for sale to the public that allows you to get in, speak an address, and do nothing else for the entire journey until you climb out of the car at your desired destination. I’m not confident about this happening before 2025.
Too late for an ETA, but I was going to address this too. Mesh networks rely on devices talking directly to each other rather than having a central point playing relay between everyone. So the cars will be talking to each other and eventually to transmitters set up as part of the local infrastructure to communicate local conditions. i.e. That pothole was filled.
Yes, network security is going to be a nightmare, as well as privacy issues, but those are conversations for other threads.
It’s not the toughest technology on earth to implement. The problem is getting government to pay for it. Just a cheap fill and run job isn’t that hard but the localities don’t want to dedicate the resources for that. One solution may be to require the self-driving car companies to pay for this somehow, which would just be a tax on the cars, and the companies will pay the politicians to keep that from happening. Maybe once enough self driving cars get wrecked something will get done.
What other thread? I’d be curious about the arguments.
You seem to be implying that commercial detection is an inherently easier machine learning task than self-driving cars and therefore the latter will never happen if you need a perfect AI for the former.
I think both of those assumptions are incorrect. It may very well be that commercial detection is harder than creating self-driving cars given that commercials and tv programs put out very similar audio and visual. My guess is you could probably detect mentions and conspicuous framing of products and/or urls/phone numbers and detect commercials fairly easily, but intuition on what’s easy is often mistaken.
Let’s bet! Do you think we’ll both still be on the Straight Dope in 10 years? 15 years?
How about we place a friendly wager of SD membership dues for life on whether there will be self-driving cars available to the public and reasonably widespread. I’m willing to be pretty flexible about the definition of widespread too. Like, 25% of the vehicles on the road being fully autonomous would qualify as widespread to me.
You have many of the big car companies working on self-driving technology, successful trials of self-driving cars in a number of cities and the only objection being generally poorly-thought out What-ifs on forums like this (“What if the car is doing 90mph around a blind bend on a mountain road, and then Cthulhu is round the corner?”)
It’s going to happen, and soon, the only thing we can really dispute is the exact timeframe, and whether the whole infrastructure will be suitable for self-driving cars or we quit at, say, 98%.
I saw a picture somewhere of a man in the drivers seat of his self-driving car, the car was in traffic, and he was sound asleep. This image cemented the inevitability of self-driving cars for me.
(Hopefully that photo wasn’t fake lol)
I’ve been driving for over 50 years, and have hit countless potholes in that half century. But only once has that been bad enough to cause an almost immediate flat tire, which might have been dangerous. But even then, I was able to move to the far right lane and down an exit without any accident. A lot of pothole-hits may have messed up the alignment, or decreased tire life or similar increased wear-and-tear on the vehicle, but only that one could have been an immediately unsafe condition.
Meanwhile, at least a half-dozen times, I’ve been in situations where someone changing lanes suddenly has caused an accident or a near-accident.
Seems like hitting the pothole, even at the risk of damaging the vehicle, is safer than suddenly veering to avoid it.
I’ll bet as well. I say that there will be less than 10% self-driving cars on the road (driving themselves, not cars that are CAPABLE of self-driving) in the United States (including Alaska and Hawaii) by 2025.
Pretty good bet on that one. There’s just not enough time to work out the bugs in the cars and develop the regulations required in just 8 years. People aren’t going to throw away their working cars to get a self-driving one when they become available either. There will be a corner to turn at some point though, and it will only take a few years for self-driving cars to hit the 50% mark after that. That’s just private passenger vehicles though, commercial vehicles will follow a different course depending on how well the technology develops and the comparative costs.Taxis and small delivery vehicles may take the lead there while large trucks lag behind due for safety reasons and the associated liability costs.
Do you think we’ll both still be on the Straight Dope in 10 years? 15 years?