This is my biggest concern with a massive, computer controlled system. The world is full of bad actors and without a doubt some of them will try and fuck over any automated system. And when you consider the computer safety of infrastructure, it does make me concerned. And I say this as somebody who would love to ride in a Johnny Cab sometime before I die…
I concur with this. Tracked roadway systems are the future solution.
The interim between here and there will be messy and combative.
The same problems occur with old fashioned human driven cars though, it’s not a new drawback for self-driving cars. There are plenty of existing drawbacks though.
As noted above, the communications won’t be direct car to car, there will be a network, likely with multiple transceivers operating along the road.
How exactly would that work? I mean, what’s the actual mechanical process for the cars interacting with the road and responding?
Life size slot cars?
Good thing none of those cars ever end up flying off the track into a ditch.
Sometimes that was the fun part!
A start is replacing visual queues (road signs, lane markers) with GPS-driven RF systems. Cars trying to gather visual information like human drivers do is pretty inefficient.
I think people underestimate the massive cost involved with doing this, though. Personally, it doesn’t seem workable.
Actual industrial systems that do these kind of things use WiFi.
Yeah, I probably didn’t phrase my question clearly. I’m trying to understand the whole infrastructure and process of a tagged roadway - what has to be embedded/available, and what is the car reading, interpreting, and doing?
The road should provide the car with all information that would be open for interpretation. exact lane configuration, all information road signs, observed other users, traffic light status. All ambiguity should be removed. The car knows what lane it should be in, at what speed and where that lane is.
For a road that has other users than just cars/trucks such a system will be insanely complex, but still is an easier problem to solve than getting a (every!) car to interpret the same road conditions with much more limited information.
So, each car would be continuously calculating its next action, taking into account all the vehicles around it and their instantaneous reactions as well? Or is the system passing continuous instructions to the car, that the car continuously follows?
I’m not @The_Librarian, but I’ve often taken the same POV.
Today the road is responsible for having adequate signage, lane markings, curbs / kerbs, traffic signals, etc., so the human intelligences steering each car know what is on offer. And the humans drive (mostly) in accordance with the information the road offers to them. These are all cues designed for human visual recognition and the strengths and weaknesses of that recognition.
I suggest the better way to do SDVs is to have SDV-capable roads serving up the exact same info, but in ways that are most reliably and unambiguously communicated to the machine “intelligence” driving SDVs. Which would be some standardized secured electronic broadcast networking. Not “machine vision” aimed at recognizing the unaltered human-oriented signs.
It gets you distance, which is something that cameras have to infer. Pretty big deal to have a second source of that information.
I would say that the next logical step is to put the machine “intelligence” outside individual cars. Then you can optimize traffic lights, anticipate for conditions miles ahead, clear the road for emergency vehicles, use counter flow as conditions demand and all kinds of other cool and efficient stuff.
Assuming that millennium is reached when every motor vehicle is a SDV, the human-oriented signs and signals are still going to need to be there for the pedestrians, cyclists, and motorcyclists who’ll also be using or crossing the roads. And the road system better be fully aware of their locations and trajectories.
Because all those signs are there to inform pedestrians about the speed limit and such?
If you try to build complex system of cars talking to each other you introduce a whole lot of complexity and potential network effects you didn’t anticipate. This is a dead end, imo. Self-driving cars need to operate autonomously. You can have aids that help, but the car has to drive safely without them. And the designers have to assume that things will happen on the road they couldn’t possibly anticipate in advance, and the car will have to respond to them logically and safely. Trying to make the environment perfectly predictable is a non-starter.
Even trains are hard to network, and they drive on fixed tracks and report their location constantly. Even so, automated switching systems have caused accidents by putting two trains on the same track at the same time.
In a complex environment, decisions have to be made from the bottom up. You can’t centrally control them. It’s fine to have a system to give cars some traffic information, weather and road conditions, etc., but the car has to be able to operate safely without them.
Motorcyclists should know the speed limit and all that. Bicyclists would also find some of that stuff handy. All road users need to know who’s expected to stop and who isn’t at an intersection.
I would think that mixed use roads (cars, pedestrians, bicycles) would be late adopting self driving.
It might be because where I live (Holland), that I don’t see self driving cars share the road with bicycles.