Cons and pros of a network controlled infrastructure system with AVs?

[Quietude]

Originally Posted by lvmensch

Actually nothing wonderful about Mountain View or Prescott. Similar to any newer suburbs. And SF is tough. [...] I would expect the rest of the nation to go reasonably fast once they can operate fully in the SW.

All of these areas have new, well-maintained streets, curbs, lane and edge markings, crosswalks etc. Whatever California's other faults, it does have road signage, markings, traffic lights etc. raised to a high art with great consistency of implementation. The Arizona cities are basically clones of that model.

Anyone who's lived outside these fully modern city/suburban landscapes knows that most states and cities have nowhere near as good a road infrastructure, even in relatively new, well-maintained areas. There are many interchanges and road combinations here in the Denver area that are quite new (under 10 years, most of them), completely conform to state and regional codes and regs, and are mind-bogglingly confusing and difficult to navigate even with experience. (And when crowded with other confused drivers.)

And then there is the other 80% of US roads where you're lucky to have a clear centerline, never mind edge markings or curbs, streetlights, traffic controls that conform to modern standards, or anything like adequate warning and guidance signage.

I used to live in a small Northeastern town. Roads were "pretty good" by most standards. I drove about six miles to the post office, and could do it without much conscious thought. I doubt any AV currently in existence could do it without considerable trouble - faint centerline, ragged edges on swervy country roads, no signage, sudden turns, unmarked intersections (all products of antique, grandfathered waivers to modern regulations). You could not see most traffic lights once you were at the limit line - tiny intersections, high wire-hung lights that were beyond the upper windshield line.

Toss in some snow and further obscuration of markers and road edges...

Current AVs are fine... on the very best of roads. Most of the time. They have a long ways to go before they work outside of Google's parking lot. But like most projects driven by tech weens with incredibly narrow understanding and processes, cars that can lap the manicured tech-campus lot without hitting the boss's Mercedes must be ready for global duty...

[lvmensch]

Originally Posted by Quietude

All of these areas have new, well-maintained streets, curbs, lane and edge markings, crosswalks etc. Whatever California's other faults, it does have road signage, markings, traffic lights etc. raised to a high art with great consistency of implementation. The Arizona cities are basically clones of that model.

Anyone who's lived outside these fully modern city/suburban landscapes knows that most states and cities have nowhere near as good a road infrastructure, even in relatively new, well-maintained areas. There are many interchanges and road combinations here in the Denver area that are quite new (under 10 years, most of them), completely conform to state and regional codes and regs, and are mind-bogglingly confusing and difficult to navigate even with experience. (And when crowded with other confused drivers.)

And then there is the other 80% of US roads where you're lucky to have a clear centerline, never mind edge markings or curbs, streetlights, traffic controls that conform to modern standards, or anything like adequate warning and guidance signage.

I used to live in a small Northeastern town. Roads were "pretty good" by most standards. I drove about six miles to the post office, and could do it without much conscious thought. I doubt any AV currently in existence could do it without considerable trouble - faint centerline, ragged edges on swervy country roads, no signage, sudden turns, unmarked intersections (all products of antique, grandfathered waivers to modern regulations). You could not see most traffic lights once you were at the limit line - tiny intersections, high wire-hung lights that were beyond the upper windshield line.

Toss in some snow and further obscuration of markers and road edges...

Current AVs are fine... on the very best of roads. Most of the time. They have a long ways to go before they work outside of Google's parking lot. But like most projects driven by tech weens with incredibly narrow understanding and processes, cars that can lap the lot without hitting the boss's Mercedes must be ready for global duty...

i do not think you understand how these things navigate. They use a higher precision GPS and a mess of other sensors. It knows where they are within a quarter inch or so. And they can use a combination of inertial and optical to maintain precise location even when the GPS vanishes.

And each device reports anomalies immediately. Road cone goes up first one who sees it reports it and anything else relevant.

i would also expect that at least the early versions will feed a maned center when anything odd happens. Reasonably easy to provide access to the sensors remotely so a central operator will be able to intervene if needed.

And there are lots of suburbs in the SW which have limited infrastructure. My neighborhood in reasonably good Las Vegas has no road markings, curbs, sidewalks or streetlights. And these guys are operating in San Francisco which is almost as bad as it gets in the US.

So yes they are avoiding snow and rain rich places. But not congestion and ugly infrastructure.

And I do not think the navigate part is even a challenge. The dynamic and rare cases are what will drive them batty...the thing that occurs every billion miles.

[Quietude]

Originally Posted by lvmensch

i do not think you understand how these things navigate. They use a higher precision GPS and a mess of other sensors. It knows where they are within a quarter inch or so. And they can use a combination of inertial and optical to maintain precise location even when the GPS vanishes.

Ah, good. It should have no trouble with the areas where the GPS screams at the driver for driving in the river 100 feet from the actual road... which is legion in the rural northeast. The mapping there, and in many other areas of the US, is somewhere below terrible. (And that's on the long, empty straights. Never mind the tangled web of a five-road intersection with two stop signs.) But it's to the millimeter in San Jose, so that's good enough, right?

And I do not think the navigate part is even a challenge. The dynamic and rare cases are what will drive them batty...the thing that occurs every billion miles.

I think this falls into the area of grossly underestimating the real-world complexity of the problem. Much like the grueling problems of Tesla as a manufacturer, the evolution of a genuinely reliable AV has issues that no amount of app-writing will fix. The vague pie-from-the-sky notion that gigadata will solve all is just... so Silicon Valley.

[lvmensch]

Originally Posted by Quietude

Ah, good. It should have no trouble with the areas where the GPS screams at the driver for driving in the river 100 feet from the actual road... which is legion in the rural northeast. The mapping there, and in many other areas of the US, is somewhere below terrible. (And that's on the long, empty straights. Never mind the tangled web of a five-road intersection with two stop signs.) But it's to the millimeter in San Jose, so that's good enough, right?


I think this falls into the area of grossly underestimating the real-world complexity of the problem. Much like the grueling problems of Tesla as a manufacturer, the evolution of a genuinely reliable AV has issues that no amount of app-writing will fix. The vague pie-from-the-sky notion that gigadata will solve all is just... so Silicon Valley.

Actually 30 years ago when GPS first became usable for navigation I spent 6 months trying to get Garmin to fix its charts. The approach to Puerto Vallarta MX was off by over 1/4 mile in the bad direction. Discovered they all knew it...but would not fix it because it was an official government chart and they would not change it even if it was wrong. Got even stranger a couple of years later when the put out a harbor chart that was accurate. As you approached the harbor and switched to the harbor chart you jumped a quarter mile out to sea.

However all this stuff is easily corrected. The Google charts in most of the US are good to better than a foot and it is a straightforward task to get it to an inch. And as we begin to really roll the fleet of AV it will quickly converge to a fraction of an inch. Yes once in a while the vehicle will not have good data. But it knows that. So it feels its way slowly and carefully and perhaps seeks human assistance.

Again the navigation problem is the easiest one. And a decade in it will no longer exist.

[hikernut]

In the near term the biggest impacts will be (1) eliminating driving jobs, and (2) offering people who cannot drive the opportunity to own and use a private vehicle. The first is both a pro and a con, depending on one's point of view. The second is a pro.

There is one indirect con that could result from #1. If AV taxies become cheap enough that many people give up their personal vehicle, the result could actually be more traffic congestion in the short term. More AVs driving around empty, at least part of the time, en route to the next fare. Not saying it will happen necessarily, but it's something to consider.

I expect that getting to the grand vision of 100% AV use, no traffic lights, etc... will take a very long time. I'm in my 50s and I doubt I will see this on a large scale within my lifetime. In limited circumstances, sure, but not commonplace.

All my own guesses!