Robocars As the Best Way Geeks Can Save the Planet

Brad Templeton writes "I (whom you may know as EFF Chairman, founder of early dot-com Clari.Net and rec.humor.funny) have just released a new series of futurist essays on the amazing future of robot cars, coming to us thanks to the DARPA Grand Challenges. The computer driver is just the beginning — the essays detail how robocars can enable the cheap electric car, save millions of lives and trillions of dollars, and are the most compelling thing computer geeks can work on to save the planet. Because robocars can refuel, park and deliver themselves, and not simply be chauffeurs, they end up changing not just cars but cities, industries, energy, and — by removing dependence on foreign oil — even wars. I, for one, welcome our new robot overlords." (More below.)

Templeton continues: "The key realization is that while the safety and timesavings that come from having computers as chauffeurs is very important and can save a million lives every year, a number of interesting consequences come from the ability of robocars to drive themselves while vacant. This allows them to deliver themselves to us on demand, to park themselves and to refuel/recharge themselves. On-demand delivery makes car sharing pleasant and allows the use of "the right vehicle for the trip" on most trips. Self-refueling means the people using cars no longer need care about range or how common fueling stations are, enabling all sorts of novel energy systems with minimal "chicken and egg" problems. Because passengers don't care about the range of their taxis, battery weight and cost are no longer issues in electric cars and scooters."

Re:Public transportation

[Dice]

Have you ever tried to get somewhere on a bus?

Here's what Google Transit gives me for my morning commute: link [google.com]. Travel time: 2.25 hours. 3 transfer events, involving a total of 3 bus lines and 1 BART train.

Either that, or I could drive over 237 and get to work in 20min.

Re:Wow, good job!

[cailith1970]

The AI is not so powerful. Most animals can navigate in traffic of their own kind, even insects. But no, it's no tiny project -- but it's a tractable large project.

Two problems with a lot of robot navigation systems that use visual processing are handling the differences in light at different times of the day, and handling a dynamic environment.

The environment can look very different even just comparing morning and evening, not touching night or times of the year. This makes following a path that you learned under one set of conditions look like it's a different path in another set of conditions. It's not a problem in indoor environments that have controlled lighting, but in a "real" scenario, it's not a toy problem.

Re:No.

[btempleton]

Is diesel more fuel efficient? You get more miles per *gallon* of fuel but isn't that just because diesel is denser, and you have more pounds of fuel -- which is what really matters -- in each gallon. It's why the gallons cost more.

Reverse for ethanol, where a gallon has only 75% of the energy of a gallon of gasoline.

Only in the US

[tknd]

While he does make this conclusion about U.S. data, he is fair and continues his search to other parts of the world like Europe and Asia. From this page [templetons.com]:

Don't Europe and Asia do better?

Much better. This Australian Study cites figures saying that Western Europeans use only 76% of U.S. BTUs/pm in their private transport, and only 38% in their transit -- 2.5 times more efficient. Rich Asians do even better at transit -- they are almost 4 times as efficient in terms of energy/passenger-mile.

Re:Public transportation

[kesuki]

"One word. Mass. The mass of a battery pack is likely to be hundreds of pounds. (The Tesla roadster's battery weighs 900 lbs, for a point of reference.)"

http://en.wikipedia.org/wiki/Forklift [wikipedia.org]

'The Counterweight - is a heavy cast iron mass attached to the rear of the forklift truck frame. The purpose of the counterweight is to counterbalance the load being lifted. In an electric forklift the large lead-acid battery itself may serve as part of the counterweight.'

unfortunately wikipedia doesn't mention they weight of electric forklift batteries, but IIRC they are 1000-2000 lbs, and are completely modular for replacement while the batteries charge.

it's not hard to do, although on forklifts it requires a pallet jack, and one human to manually service a battery. my point was that for this to be 'realistic' for a cab company, the battery MUST BE MODULAR, even if a human with a pallet jack replaces the battery instead of a big robot.

if a large lead acid battery the size of a forklifts can be modularized, despite being a very big thing full of lead and acid, then a robotic cab can modularize the battery, as well.

Re:Public transportation

[ryanov]

I love that you're calling someone else retarded while simultaneously saying something that makes no sense. Buses don't go near people's homes, require tons of fuel, have inconvenient times, and don't allow cargo transport, you say. Funny, the house I grew up in, in the suburbs, has 2 buses within 5 mins walk and each of them runs about every 20 mins depending on the time of day. Many of the buses have storage racks or under-floor storage. You say these things like they're an inherent physical property of "the bus", not bad planning or design. Our biggest problem is selfish fat fucks that don't want to do things like ride the bus, and try to invent 500 reasons that they "won't" work when they will, and do every day for plenty of people.

A more realistic view

[Animats]

Having run a DARPA Grand Challenge team, I've been through most of this line of reasoning. I'm rather less optimistic.

First, Templeton writes "The cost of accidents is arguably the single largest component of the per-mile cost of driving a vehicle", but doesn't provide justification for that statement. Total US gasoline consumption costs about $600 billion per year. The American Automobile Association says that US auto accidents cost about $164 billion per year. [cnn.com]

Second, while we can do automatic driving in a situation where all the players are reasonably well-behaved vehicles, we're a long way from being able to do it safely in a populated area. Today's robot vehicle technologies have minimal "situational awareness". That's one of the hardest problems in AI. Right now, sensing systems are up to recognizing "obstacle" and "moving car-like thing". Pedestrian and bicyclist behavior prediction is a ways off.

The whole section on robot vehicles with incredible evasive ability is bogus. Vehicles are limited by inertia and maneuvering room. Cutting the reaction time from 500ms to 50ms would help some. Half of all collisions would be prevented if braking started 500ms sooner, according to a Mercedes study. Chain collisions are an artifact of human reaction time; with minimal inter-vehicle coordination, all the cars in a lane could come to a fast stop without colliding. But evasive action requires room.

Most of the estimates of huge savings come not from automatic driving but from electric cars. Especially little lightweight electric cars. You can get little electric cars now; I'm in Silicon Valley and I see them now and then. But they're about as common as Segways.

Zipcar [zipcar.com] indicates that the car sharing concept can work. With automatic driving, the car could be delivered to you, so it could be used in less-dense areas than central cities. But it's really for people who only need a car occasionally. Zipcar is $10/hour.