While I’m sure some interesting feedback loops will crop up, let’s again not forget that humans are subject to the same. The common occurrence of a persistent slowdown in traffic (long after the initiating incident is over) is one obvious example of a case where local behavior leads to more widespread consequences. People don’t start slowing quickly enough, nor do they accelerate quickly enough when leaving the area. So you get a persistent bunching of cars until the density gets low enough that it diffuses naturally. Self-driving cars should be able to avoid that problem, at least.
A simple example: A car is designed to check for reverse lights in the row behind the car when backing put of a parking lot stall. A good safety practice. Then another self-driving car parks behind it. Both try to back out, see each other’s reverse lights, and wait - indefinitely.
Weird effects might happen in congested traffic if enough of them are self-driving.
Here is something from The New Republic arguing, among other things, that all of the talk about self-driving cars is to distract attention from mass transit projects.
The California High Speed Rail project benefits from any purported distractions. Anyone that’s paying attention can see that it’s dumb as shit to spend $128 billion on that link.
Maybe the purveyors of mass transit megaprojects should figure out how to reduce their costs and not lie about them in the first place. They can do that regardless of what anyone else is doing with self-driving.
that’s interesting/weird … even in google maps I get “in 200m use any of the left two lanes to turn left” … sounds VERY basic to me …
oh paleeeeese … anything north of the rio grande has to be considered “child’s play” … go further down into LatAm … where they steal complete locomotives (not necesarily thinking it through)
good news is - the police managed to backtrack it (after 2+ months)
Al128 is touching on something important – do self-driving cars have to drive EXACTLY like a human drives? Or is it acceptable for the self-driving car to solve the same driving problems in ways more suited to an electronic brain?
Wait until they try to spend all that money on new Amtrack lines the infrastructure bill. Especially with shortages of material and people. That’s going to be gigantic boondoggle, but not as bad as the California HSR in percentage of waste terms.
As for turning lanes, does anyone know what self-driving cars do if they need a lane and no one lets them in? Around here it’s not uncommon that when you need to get into the lane for the exit of a freeway, sometimes people will literally cut you off if you signal first, filling up the gap beside you. Or, the lane you need will be quite crowded, and you either have to accelerate hard to get to an open spot or slow down below the rate of traffic for a hole to open. Failure to do so means missing your exit.
So what does a self-driving car do if it can’t make a safe lane change into an exit lane? Just miss it and try for the next exit? Because I expect other drivers will do that sort of thing a lot more if they don’t have to worry about retaliation from the other driver.
The approved Amtrak funding is ~$4B per year for the nation. Spread out, I think that’s a reasonable level of funding. I’m not against rail, or maintaining infrastructure. I am against these megaprojects that are mismanaged at an incredible scale. California HSR is one of these.
What I especially oppose are megaprojects that are not progressive in their utility. California HSR is useless until it connects LA with SF. But general Amtrak upgrades benefit local riders immediately.
I’m fairly happy with the local BART (Bay Area Rapid Transit) upgrades, for instance. They slowly inched their way down the East Bay until they connected with Milpitas (near where I live). Each new station gave the system more utility. But between north and south California is a wasteland.
Well, they certainly aren’t going to just shove their way in. I guess that just means they’ll miss the exit.
One advantage of a no-LIDAR system is that it’s a lot harder to tell if there’s a driver or not . Maybe I should get an inflatable dummy for my Tesla:
I’m being hyperbolic obviously, but at the same time I’m not kidding that almost all the value in the system is connecting the Bay Area with LA. That’s also how it was sold to the public.
I recently saw an interview with one of the HSR project leaders, and they defended an incomplete link as still benefitting disadvantaged communities. Fine, but those communities are small, and it would be cheaper to just buy them free cars and gas for life. And again, that’s not how it was sold to the public. This is an after-the-fact justification. No one would have approved tens of billions of dollars just to connect to Madera.
With respect to rail, California has a major disadvantage in that there is very little mass transit available to get you to and from the train station.
San Francisco has the best transit infrastructure in BART (It’s a great way to get from SFO to the City), but if you consider other cities in California, they are pretty dicey with respect to convenient transit option once you get there by train.
Los Angeles has both light rail and a subway, which are almost always guaranteed to be able to take you directly to just about anywhere except where you need to go. San Diego has a trolley! Both cities are large, geographically (especially the Los Angeles metro region, which is larger than many states), so this is sort of a big deal.
L.A. has been incrementally expanding our light rail systems. It’s a generational project, but it is getting better. There are also more dedicated bike and bus lanes.
That part it gets, but a (good) human driver may move over at 1000-200m to take advantage of a gap in traffic. If FSD plans to move at 200m, it won’t move until 200m, even if the lane is clear at 300m, but backed up at 200m.
I’m not really sure what it will do if it gets blocked out, as I’ve not been willing to let it.
The real problem is when it moves into a turn lane even though the route goes straight. Or takes a wrong exit, or other things counter to what the map shows.
I sometimes switch lanes, b/c my celphone tells me the next turn is a left turn … even if I have - say - 800m to go … for the very reason … to not have to sweat it in the last 100m
Yeah, you would think that if the car knows it needs a lane at a reasonable time in the future, it would move to that lane as soon as it was clear to do so. I’m not sure I get the reason for the 200m setting. That’s awfully close for me anyway. If I’m lane changing to get into an exit lane that’s only 200m away, I did something wrong. At least if the road is congested and there’s a chance the lane will be blocked closer to the exist. Why push things?
The lane changing problem is one I’ve used to describe why self-driving cars won’t ever work the same way that humans are used to driving:
There’s one physical thing which makes lane changing easy for humans, but impossible for computers: eyes.
People often make eye contact with the person next to them, and indicate who makes the next move. (at 4-way stop signs, at pedestrian crosswalks, at intersections, at merging lanes, etc)
When we all get used to driving using zero eye contact, then we will feel comfortable sharing the road with autonomous vehicles.
. Maybe I should get an inflatable dummy for my Tesla: