Comment by robomartin | Hacker News Reader

Comment by robomartin | Hacker News Reader

Parent Full thread

> How do self-driving cars reduce congestion?

At what scale? As a non-trivial percentage of road traffic they can reduce congestion in a million different way. For example, predictive and cooperative route planning (you know what the road looks like 0 to N Km ahead and plan accordingly.

The other one is one of the most frustrating things humans do while driving: Slow down while going uphill. Where I live, except of accidents, this is probably the #1 source of traffic jams on the highway. I don't know what it is with some people (and it doesn't take but a few to cause a mess). Are they not paying attention? It's a chain reaction, once a few cars slow down, traffic coming up at speed log-jams behind them. And, of course, because in the US (or, at least, in CA) nobody has the courtesy to move over the right if they want to go slow, you easily end-up with 4 to 6 lanes of perfectly good highway absolutely jammed. Once over the crest every seems to drive like Max Verstappen. It's both tragic and hilarious to observe.

So, yeah, self-driving cars --at scale-- can absolutely reduce congestion by driving far more intelligently than most humans do.

That said, at a scale that approximates a rounding error they probably can't materially affect traffic.

The average passenger capacity per mile of a Waymo is <1. The average for a regular passenger car is >1.

Moving the same number of people with fewer passengers per car increases traffic.

You're not factoring in the unused capacity of a regular car parked on the street or in a garage versus being in service of delivering transportation value..

The point of this idea is that housing is expensive to build and one way to significantly reduce that cost is by eliminating parking. Add to the cost of car ownership and the cost of living is substantially higher.

Most people still will need a car to get around. Bus rail and carpooling are not real solutions for most people in urban USA cities. Some feel offended by this notion, but it's reality.

I imagine that the concept of cities leasing waymos and being able to allocate them around to particular nodes will increase the avg. passenger capacity as well, because it can be modeled and tracked to figure out where the nodes need to be to create an effective network with high utilization.

But yes, traffic is indeed an impact.

The government can also lease vehicles with more capacity, like mini-busses if that's the policy objective. Doesn't have to be cars. The government would be in the driver's seat to drive this driverless policy as well, no pun intended

An unused car parked isn't causing traffic jams on the highway. It's irrelevant to my comment that AVs increase traffic, not decrease traffic. I agree reducing parking needs is a benefit to AVs, but arguing it will reduce congestion is ignoring extremely basic traffic concepts.

You've got more vehicles on the road driving to move the same number of people. What does that do to traffic??

The actual savings are minimal at best, as demand for cars is all about peak, and how the peak travels typically in similar directions. You see this in trains. In downtown SF, you will find trains parked near the stadium all day, because there is no,practical use for them 9p% of the day. Most of the fleet is idle most of the day, and it would not be any better if it's a self driving car. You'd have a much better chance if trips were random, but they absolutely aren't.

The nodes have been calculated and modeled, the gains round to zero. What you get instead is higher demand for moving around in cars, because it's now cheaper and less annoying. Lower prices increase demand, therefore more congestion, not less. This component obliterates said minor savings from earlier in any reasonable model.

Waymo is not doing any carpooling optimizations. They're right now a luxury service that is priced above the normal Uber ride prices.

This will obviously change in future, once they start a truly mass rollout.

Yeah, all the people who refuse to ride busses to the point they're willing to spend many many thousands of dollars on their own cars are just yearning to carpool in a much more compact vehicle than a bus.

We don't need to move _all_ the people to carpooling to fix the congestion.

And remember, the alternative is even worse: buses.

I know it's not popular in some circles to say it, but people (not just evil, rich racists) dislike riding the bus because of homeless people and other people who have little reason (or ability) to conform to social norms. I only saw one a couple altercations on the bus in my years riding it in Austin, but that's enough to get a lot of people to decide it's not worth it, especially women or people with kids.

We're not willing to enforce basic safety on buses, let alone other norms. If Waymo is willing to do that, that will make it available to many more people. Even just not having to wait at the bus stop before sunrise can be a game changer.

> Yeah, all the people who refuse to ride busses

It's not that people refuse to ride busses. In most places in the US this is either a non-existent or extremely impractical option. It's just reality.

I actually really liked uber pool in the Bay Area, talked to some interesting people

> Moving the same number of people with fewer passengers per car increases traffic.

The overwhelming majority --overwhelming-- of cars, SUV's and trucks on the road have exactly one passenger, including the guy commuting to his office job in an F-250.

You're right, but that still puts the minimum at 1. Even with the many, many vehicles with single passengers as 1, the average is still always higher than 1.

An autonomous vehicle doing a deadhead drive has a passenger count of 0. 0 < 1. You start tossing in a lot of miles with 0 in there on top of the vast majority of your trips being 1, and suddenly now you're less than the cars you're theoretically replacing.

A lot of these AVs at the moment at least are still roughly minivan/midsize SUV. On top of that, with safe distances adding another foot of the car to a highway scenario really doesn't increase the percentage of the total space that much. You need a lot of follow space ahead/behind the vehicle.

Maybe if they were like single person pods that teleported to their pickups and managed to group tightly together in packs on the road, but until then they're just another SUV on the road driving around sometimes with zero people in the car.

I don't necessarily disagree with you. I think single person AV pods would be amazing and is likely one version of the future. However, this kind of change takes time, a lot of time. AV or not, if you have to share the road with monster trucks being driven by humans who are too busy looking at pictures on their phones, well, you are going to eventually want to be in something at a similar scale. That's why interesting options like Smart cars never made it here.

Whatever this will become, it will be a matter of decades and a real modernization of our traffic laws and other infrastructure. I will not pretend to know what form that will take, my crystal ball stopped working a few decades back.

There are so many questions. Just looking at a place like Los Angeles, I regularly drive right next to massive trucks moving containers and all kinds of cargo. Can we be more intelligent about this? An automated cargo monorail operating down the middle of every highway? Sure, as an idea and a fantasy. In the real world you'll have to contend with myriad issues spanning a range between our inability to build anything anymore in this country with how you are going to deal with millions of people losing their jobs if you automate cargo transport and clear large trucks from the highways. Etc.

It's nice to have these visions about the future. What will actually happen is a different matter. I am placing my bet on self driving cars as the first practical phase of a transition into something more sensible that will likely require three to five decades. The equation is very simple: Atomos are expensive and hard to move. Conventional cars and trucks manufactured and purchased today are going to stay on the roads for at least a decade. A mass transition to a different paradigm is measured in units of decades after the last conventional vehicle is manufactured and sold. Hence my five decade estimate for transportation utopia.

I wonder what the logistics and costs look like having to maintain a fleet of single-seaters along with multi-seater vehicles. I realize a high percentage of rides are solo rides, but there's still like 20-30% of rides with >1 passenger. Is it really easier for the transit companies to manage multiple fleets of vehicles with various seating sizes? Is it actually a better experience for riders?

> An automated cargo monorail operating down the middle of every highway?

There's already cargo freight rail practically all across the country. Loads on trucks are on trucks because of the desired flexibility and point to point response time as opposed to having to load/unload rail.

Also, unless you're talking some really long distances, having rail in the middle of a highway doesn't always make much sense. You're not going to have the depots or stations in the middle of the highway after all. The train then needs to weave back and fourth across the highway to get to the places its routed to. Far better to just figure out its own right of way rather than stick it in the middle of a complicated place.

> Once over the crest every seems to drive like Max Verstappenit's odd, you'd think they'd all try to be ahead at the apex

I drive track days with some frequency. When I am on the road I have to turn my brain down to slow and just let it happen. To be fair, you can't expect grandma to drive at a high level. What most people talking about congestion and traffic don't stop to think about is that the traffic you get is a function of the lowest common denominator. All you need is one small rock in a stream to disrupt it.

Reply on news.ycombinator.com