Share good ideas and conversation.   Login, Join Us, or Take a Tour!
comment by veen
veen  ·  431 days ago  ·  link  ·    ·  parent  ·  post: Self-driving cars can't cure traffic. Surge pricing can.

I actually just mailed with wasoxygen about this today.

    "[Induced demand is] the great intellectual black hole in city planning, the one professional certainty that everyone thoughtful seems to acknowledge, yet almost no one is willing to act upon."

Congestion charging - the traffic planner's term for it - actually works like a charm. It's probably the only reason London isn't gridlocked all day every day. It depends on the implementation, and in US cases, it's usually toll roads and not congestion zones. Charging on roads has been shown to be effective for that road, but not for the surroundings as traffic diverts more than it reduces.

That said, I do think that AVs can alleviate traffic if they will end up being shared taxi-like vehicles. Those can be regulated and congestion-charged to hell and back. Here's the thing about traffic jams that people often forget: the distribution of vehicles over time is just as important as the amount of vehicles. Highways are built for peak demand, but never have those traffic intensities outside of the short morning and afternoon peak. (Unless you're in LA, but that's a different story.) Spreading that peak out could prevent jams could be enough to seriously reduce jams. And who knows, maybe society can finally get rid of the artificial 9 to 5 window where all business must occur.

kleinbl00  ·  431 days ago  ·  link  ·  

Having been in London, and having been in, like, the United States, I don't think an appropriate zone can be created with American sprawl.

And having done Lyft Line more than a few times, our current implementation of rideshare fucking sucks, too. You save like $2 and you're halfway home and then the driver suddenly makes a 90 degree turn and sends you squirreling 15 minutes into the Hollywood Hills and by the time you're there whoever fucked up your ride has decided they're not hungry anymore so now you're on Fairfax somewhere and the driver is lost and Lyft's GPS and his phone GPS are saying different things and if you hadn't been a cheap prick attempting to save two dollars you would have been home half an hour ago.

And then there's the times that you pick up someone on their way to the emergency room. And allow me to share the fact that someone taking Lyft Line to the emergency room is not only having a shitty week, they're also contagious.

These are all great ideas, but the implementations tend to suck.

goobster  ·  431 days ago  ·  link  ·  

We already have HOV or Carpool lanes.

I figure eventually we will just convert those to Autonomous Vehicle lanes, or build a new lane for them, like they do with buses.

This gives the AV's a "safe" environment in which to travel, they can follow each other much more closely than other cars, and they can have their own exits that don't need stupid-human-tools like traffic lights, etc.

I can't wait. I'm tired of sharing the roads with assholes. I just wanna get in, listen to a podcast, and arrive at my destination.

veen  ·  430 days ago  ·  link  ·  

Isolated high-speed tracks are the perfect environment for AVs. The control part of cars seems to have been figured out by now with almost millimeter precision. I thought however that, in general, of those lanes did not have special exits? I vaguely recall drivers in LA crossing something like five lanes to get from the carpool lane to the exit.

goobster  ·  428 days ago  ·  link  ·  

Right. The Exit/Onramp problem is simply solved with AV's, because you do not need traffic lights and 8-feet of extra space on either side to accommodate idiot drivers, and you can have walls and tunnels and narrow passages because you don't have to worry about people panic-braking because the lane got narrower, etc.

Essentially, if we had SmartCar-sized AV's, we could fit TWO lanes, side-by-side, in the space currently taken by ONE lane of idiot-operated cars.

Every single road is something like 80% wasted space and design, to account for human error, and the peculiar foibles of our sensory systems. When you eliminate those vectors of uncertainty from the calculation, you can fit FAR more vehicles into the current traffic pattern, with little/no impact on the current traffic flow and control mechanisms.

Let's look at a detailed example...

Off-Ramps and AV's

Because AV's are smaller and narrower than a regular car, you have TWO lanes of AV's operating in your current HOV lane.

Off- and On-ramps are on the LEFT side of the freeway, and ONLY accessible to AVs.

They do not need to be built to support a 40-ton semi tractor trailer. They need to support a series of 1800-pound AVs. The off-ramp also doesn't need to be flat, or provide forward visibility, or a clear line of sight, so angles, turn radii, and crash zones don't need to be designed into the off-ramp design. (Imagine AVs turning off the freeway into a simple tube, that bends down away from the freeway in a sweeping arc, towards the road below.)

The outside of the tubes can become public art spaces... paint them like snakes, or dragons, or whales, or whatever.

Below the freeway, the tube merges with regular traffic in a standard merge lane, or - more likely - adds a lane to the road that is dedicated to AVs.

AVs enter the freeway the opposite way, entering the freeway from the left, and merging in with other AVs.

Because all the AVs are talking to one another constantly, the flow of traffic in these lanes is smooth. All the vagraries of acceleration, object avoidance, jammed up off-ramps, etc, are spread amongst ALL of the vehicles in the area, so each one backs off an inch (for example) and the cumulative effect over 60 vehicles is enough space for a new vehicle to merge in.

With two AV lanes, the left one is the slower lane, with traffic entering and exiting the freeway, and the right lane is the "fast" lane for vehicles going further.

The AVs essentially become person-sized train cars, wirelessly connected, which can dynamically "connect" and "disconnect" from the cars in front and behind them, and flow in with vehicles in the other lane.

The Fun Stuff

The best part is once people get past their hangup on the "Trolley Problem" strawman argument, and the software designers get to begin thinking of how to make things flow like water, rather than road designers having to design to accommodate that Escalade with one person in it, who is also on the phone, and riding on bald tires with brakes that should've been serviced 20k miles ago.

What if an AV gets a flat tire? I could see the AVs in front and back "teaming up" and sandwiching the one with the flat tire, and helping it get safely to a turnout or Repair-de-sac (something I just made up: a little turnout that can fit 3 AVs... the "bad" one, the "replacement" one that picks up the rider immediately and continues their journey, and the Repair vehicle that carries the dude who fixes the broken one.)

When we get to this point, London reduces the number of vehicles allowed in The City even further. Taxicabs in two lanes, AVs in their own four lanes. That provides the financial base and incentive for the AV companies to develop and built resilient systems, and tweak the code in a live environment, before rollout to progressive cities worldwide... Sao Paolo, Vancouver, Toronto, Dubai, Stockholm, Berlin, Nagoya, etc.

veen  ·  425 days ago  ·  link  ·  

So I agree in the general points you make, but disagree with the specifics, and I think it mostly has to do with this assumption:

    When you eliminate those vectors of uncertainty from the calculation, you can fit FAR more vehicles into the current traffic pattern, with little/no impact on the current traffic flow and control mechanisms.

While AVs are much more accurate and can be far more stable than regular cars, I don't think AVs can eliminate those uncertainties. I also don't think that infrastructure will be changed all that much to AVs. Sure, we can redraw the lines to fit three vehicles next to each other instead of two, but one of the reasons I think AVs are such an incredibly promising technology development is that you don't need to build new infrastructure. No rollercoaster-like tubes necessary.

(Sidenote about that: you're forgetting the impact G-forces have on the rider experience. If I'm in an AV, I don't want to spill my drink because my car decided to go off the highway. The general public does not like even minor G forces, especially not in day-to-day vehicles. )

Another thing that you ought to keep in mind is that infrastructure changes will always involve or depend on the government. Seeing the sluggish progress on something non-physical like the regulatory frameworks indicates to me the government's reluctance to radically change its infrastructure to this new tech.

In my opinion the real efficiencies are to be found at the use of resources and energy to transport all the people in, say, a city. Moving to all-electric, demand-driven taxi-like AVs will, without

    Because all the AVs are talking to one another constantly

That is still something I need to see to believe. I think that vehicle-to-cellular-to-vehicle communication will be inevitable (e.g., communicating roadworks quickly to all vehicles) but that vehicle-to-vehicle communication and harmonization of control is something that will end up not being reliable enough. Platooning - the term coined for train-like AV's - is now being tested at at least a few feet of distance, and while I do see vehicles being close together based on camera vision / radar, I don't think the swarm-like AV behaviours that people dream of is actually feasible.