Ian Bicking: the old part of his blog

Re: On the subject of mass transit

Well you've almost got it. I'd make a few changes. First, I'd do away with the rail and make the cars drive themselves on the existing roads. This gets rid if the big infrastructure building problem and makes the trip door to door. Sure cars can't drive themselves yet but they're getting there. They don't have to be perfect - just better than a human being. That's a pretty low standard to meet if you ask me.

Second, get rid of the fixed routes. When you need a ride, you punch a button on your GPS enabled cell phone and tell it where you want to go. A big server farm puts that data into the system and then matches you up with an existing car that's going to the same general area as you and that also has space available. Your ride gets selected out of the existing cloud of vehicles currently on the road.

You could do this now by using drivers until the self driven vehicles become available. To scale the system for rush hour, you let people use their existing cars in a dynamic carpool. They're instructed by the same cell phone on who to pick up on the way to and from work. For off ours, you keep a certain number of vehicles availble. You'd loose money on the off ours but it would be worth it because the system works best if you can eliminate the personal vehicles.

Cars cost a fortune. By using the existing infrastructure and eliminating the need for a personal car, you should be able to move more people much more cheaply than any existing system. The only problem would be writting the software that matches people up to the best car for their trip.

Comment on On the subject of mass transit
by Joe Goldthwaite

Comments:

That's very much like the rail option, except at-grade. This works okay as long as the system is closed. That is, people are not allowed in the system. Being better than humans at working with humans is a nearly AI-complete problem. Judging, for instance, the intention of a biker or pedestrian at a crosswalk requires a subtle reading of body language. So, get rid of the humans and you are all set.

Of course, this kind of environment is being set up in many communities. Many suburbs are already human-hostile, with more and more "order" being imposed with special lanes, traffic signals, etc. But it doesn't appeal to me very much. The technology for that kind of automation is crazy hard.

A more modest -- but still far from implemented -- goal is cars that are automated on the highway. In this model you have a closed system -- special highway lanes -- that cars can reliably work in. The infrastructure costs are lower because it isn't a complete system, and relies on drivers for much of the driving (everywhere except the highway). It's incremental to a degree, because not everyone has to use those lanes, though the people that do will have to replace or upgrade their cars. There is an efficiency because the lanes are much higher capacity. In theory I think there is some energy saving, because cars might actually travel close enough together to decrease drag; though I'm not sure if that's actually significant. In some proposals the cars go quite fast (e.g., 80 or 90mph) which means significantly reduced mileage. One of PRT's potential advantages is not the raw speed (though some systems are designed for high speeds), but running at a consistent and efficient speed. This is good for efficiency. The low-hanging-fruit of energy efficiency vs. speed is to increase (or at least retain) average speed, while decreasing maximum speed. Automated car systems typically only increase maximum speed to increase average speed.

# Ian Bicking

Reading my own reply, I wonder how I kept typing "off ours" instead of "off hours". My "H" key seemed to be working ok everywhere else. Brain fart I guess.

Like I mentioned, you could use drivers until the vehicles could drive themselves. I pictured a system where independent drivers would purchase their own vehicle and join the driver network. They wouldn't have set routes but would get directed by the cell phone or other GPS enabled device to where they pick up and drop off passengers. The driver network would take care of all the billing. Individuals who work other jobs could join the driver network during rush hour to help pay their personal vehicle expenses. Drivers working at night when there's less demand would get paid more per passenger. Even though the network would loose money at night, it could make it up during the day but service would be available 24x7.

With the cell phones networked together with built-in GPS, the hardware is in place to do this now. The software to track that many vehicles and match up the passenger with the best car is the difficult part. I don't even know if it's possible but it would be interesting to build some sort of simulation just to get a feel about how hard it would be. I've wanted to work on it for a long time but I'm too busy and I'm really not bright enough to do that kind of work.

If you did get it working, the other problem would be getting people to use it. By nature, it works better with the highest volumes. If you've got 10,000 cars hooked up to the network, the chances of you finding a ride going the same route as you is a lot better than if you only have 500 cars. At 100,000 cars, you probably wouldn't have to wait long for a car and the car wouldn't have to make very many stops. To get higher participation, you have the government subsidize it for a while. Companies that offer carpool incentives could use that to encourage people.

In short the problems are both technical and political but it would be nice if someone could get it working. It could revolutionize transportation worldwide. Like you said, current mass transit doesn't work very well. Here in Phoenix Arizona, we're too spread out for any traditional mass transit to work. Here, you need a car.

# Joe Goldthwaite

What you describe is actually very reminiscent of the micro bus lines common in the developing world. Coincidentally, these are often actual VW Microbuses, driving on kind of ad hoc routes and picking people up. They don't take you to your exact destination, and there's no high-tech way to manage or get a pick up, but they are definitely a more decentralized kind of bus.

It's kind of weird, but the developing world seems to generally have much better public transportation. How buses can run at a profit with $0.50 (or less) fares, I don't quite understand. But for the most part they really do seem to do well without much government help (usually fuel is subsidized in some fashion, I think). Lower wages, certainly. And their maintenance is much less... thorough. And they drive fast. And they use old vehicles -- less capital costs. Still, kind of crazy they can make it work and profit. That said, here in the non-developing world taxis are expensive. They are actually taxed, not subsidized, and they don't pick people up in the same way, and they have higher capital costs. But I worry that there are basic economics that make them unpractical. Drivers and mechanics are paid more here. People expect a nicer riding experience. But maybe it isn't impossible, if done right and smart.

# Ian Bicking

Yeah. I guess people would have a hard time paying the actual cost but that's more of a perception problem. The system would have to be cheaper than owning your own car. People are paying the cost now it's just hidden in their existing budgets.

As a little exercise, I tried to figure out about what my current per-trip cost is for my 2001 Tundra pickup. It cost 18,000 so lets assume I can drive it for ten years so the yearly cost is about $1,800. During that time, I'll figure that regular maintenance will average $500 per year. I drive around 16,224 miles a year making six 52 mile round trips a week. I get about 18 miles per gallon so my yearly gas bill should be around $2000. I also pay about $1,200 in insurance. That comes out to a total of $5,500. I'm figuring on 312 commutes to work a year so every single time I drive to work and back it costs me about $17.62. I probably wouldn't pay $3 for a one way bus ride but I'll pay much more than that to drive without even knowing I'm doing it.

That's part of the political problem. You can create a system that's much more efficient because you increase the ride density. Put an average of four people in that car without adding too many miles to the trip and you greatly decrease the per-person cost. There are billions of dollars in potential savings! Plus, if you can get enough people using it, it decreases traffic which increases both fuel and time efficiency. In many places, it would speed up commute times because even though you're making more stops, you wouldn't be stuck in traffic.

Bootstrapping the system would be the hard part. If you can't get enough people to use it, it wouldn't work. The wait times would be too long. I have my doubts about whether you could ever get it up and running but if you could, it would be the only mass transit system that would work out here in the west.

# Joe Goldthwaite

In Russia (at least when I visited in 1995-96) there are "routed taxis" (marskrutnaya taksi) and "taxis". The routed taxis are vans and just take you to the nearest metro station. They fill the gaps in the streetcar/trolleybus/autobus network in the outer parts of the city.

The "taxis" are private cars, driven by people who need money or are just bored, so they drive around all night giving people lifts for maybe $3 a trip. People who don't want to be seen taking public transportation take "taxis" instead, and everybody uses them in the off-hours. You just hold out your hand (straight out, not thumb up) at the passing cars until one stops, and then negotiate a fare. I had friends who drove them occasionally. I wondered if they were safe but I never heard of anybody getting robbed in one.

# Mike Orr