Brad Ideas

Robocars & More

I decided to gather together all my thoughts on how robocars will affect urban design. There are many things that might happen, though nobody knows enough urban planning to figure out just what will happen. However, I felt it worthwhile to outline the forces that might be at work so that urban geographers can speculate on what they will mean. It is hard to make firm predictions.

There are many fields that people expect robotics to change in the consumer space. I write regularly about transportation, and many feel that robots to assist the elderly will be the other big field. The first successful consumer robot (outside of entertainment) was the Roomba, a house cleaning robot. So I've often wondered about how far we are from a robot that can tidy up the house. People got excited with a PR2 robot was programmed to fold towels.

This is a hard problem because it seems such a robot needs to do general object recognition and manipulation, something we're pretty far from doing. Special purpose household chore robots, like the Roomba, might appear first. (A gutter cleaner is already on the market.)

Recently I was pondering what we might do with a robot that is able to pick up objects gently, but isn't that good at recognizing them. Such a robot might not identify the objects, but it could photograph them, and put them in bins. The members of the household could then go to their computers and see a visual catalog of all the things that have been put away, and an indicator of where it was put. This would make it easy to find objects.

The catalog could trivially be sorted by when the items were put away, which might well make it easy to browse for something put away recently. But the fact that we can't do general object recognition does not mean we can't do a lot of useful things with photographs and sensor readings (including precise weight and other factors) beyond that. One could certainly search by colour, by general size and shape, and by weight and other characteristics like rigidity. The item could be photographed in a 360 view by being spun on a table or in the grasping arm, or which a rotating camera. It could also be laser-scanned or 3D photographed with new cheap 3D camera techniques.

When looking for a specific object, one could find it by drawing a sketch of the object -- software is already able to find photos that are similar to a sketch. But more is possible. Typing in the name of what you're looking for could bring up the results of a web image search on that string, and you could find a photo of a similar object, and then ask the object search engine to find photos of objects that are similar. While ideally the object was photographed from all angles, there are already many comparison algorithms that survive scaling and rotation to match up objects.

The result would be a fairly workable search engine for the objects of your life that were picked up by the robot. I suspect that you could quickly find your item and learn just exactly where it was.

Certain types of objects could be recognized by the robot, such as books, papers and magazines. For those, bar-codes could be read, or printing could be scanned with OCR. Books might be shelved at random in the library but be easily found. Papers might be hard to manipulate but could at least be stacked, possibly with small divider sheets inserted between them with numbers on them, so that you could look for the top page of any collected group of papers and be told, "it's under divider 20 in the stack of papers."

The folks at the SARTRE road train project have issued an update one year into their 3 year project. This is an EU-initiated project to build convoy technology, where a professional lead driver in a truck or bus is followed by a convoy of closely packed cars which automatically follow based on radio communications (and other signals) with the lead. They have released a new video on their progress from Volvo.

Two bits of robocar news from last week. I had been following the progress of the Stanford/VW team that was building a robotic Audi TT to race to the top of Pikes Peak. They accomplished their run in September, but only now made the public announcement of it. You can find photos and videos with the press release or watch a video on youtube.

Nissan is touting that the EPA gave the new Leaf a mileage rating of 99mpg "gasoline equivalent". What is not said in some stories (though Nissan admits it in the press release) is that this is based on the EPA rating a gallon of gasoline as equivalent to 33.7 kwh, and the EPA judging that the car only goes 73 miles on its 24kwh battery.

You're driving down the road. You see another car on the road with you that has a problem. The lights are off and it's dusk. There is something loose that may break off. There's something left on the roof or the trunk is not closed -- any number of things. How do you tell the driver that they need to stop and check? I've tried sometimes and they mostly think you are some sort of crazy, driving to close to them, waving at them, honking or shouting. Perhaps after a few people do it they figure it out.

Today, I was challenged with the question of how well robocars would deal with deer crossing the road. There are 1.5 million collisions with deer in the USA every year, resulting in 200 deaths of people and of course many more deer. Many of the human injuries and crashes have come from trying to swerve to avoid the deer, and skidding instead during the panic.

At present there is no general purpose computer vision system that can just arbitrarily identify things -- which is to say you can't show it a camera view of anything and ask, "what is that?" CV is much better at looking for specific things, and a CV system that can determine if something is a deer is probably something we're close to being able to make. However, I made a list of a number of the techniques that robots might have to do a better job of avoiding collisions with animals, and started investigating thoughts on one more, the "flying bumper" which I will detail below.

Spotting and avoiding the deer

• There are great techniques for spotting animal eyes using infrared light bouncing off the retinas. If you've seen a cheap flash photo with the "red eye" effect you know about this. An IR camera with a flash of IR light turns out to be great at spotting eyes and figuring out if they are looking at you, especially in darkness.
• A large number of deer collisions do take place at dusk or at night, both because deer move at these times and humans see badly during them. LIDAR works superbly in darkness, and can see 100m or more. On dry pavement, a car can come to a full stop from 80mph in 100m, if it reacts instantly. The robocar won't identify a deer on the road instantly but it will do so quickly, and can thus brake to be quite slow by the time it travels 100m.
• Google's full-map technique means the robocar will already have a complete LIDAR map of the road and terrain -- every fencepost, every bush, every tree -- and of course, the road. If there's something big in the LIDAR scan at the side of the road that was not there before, the robocar will know it. If it's moving and more detailed analysis with a zoom camera is done, the mystery object at the side of the road can be identified quickly. (Radar will also be able to tell if it's a parked or disabled vehicle.)
• They are expensive today, but in time deep infrared cameras which show temperature will become cheap and appear in robocars. Useful for spotting pedestrians and tailpipes, they will also do a superb job on animals, even animals hiding behind bushes, particularly in the dark and cool times of deer mating season.
• Having spotted the deer, the robocar will never panic, the way humans often do.
• The robocar will know its physics well, and unlike the human, can probably plot a safe course around the deer that has no risk of skidding. If the ground is slick with leaves or rain, it will already have been going more slowly. The robocar can have a perfect understanding of the timings involved with swerving into the oncoming traffic lane if it is clear. The car can calculate the right speed (possibly even speeding up) where there will be room to safely swerve.
• If the oncoming traffic lane is not clear, but the oncoming car is also a robocar, it might some day in the far future talk to that car both to warn it and to make sure both cars have safe room to swerve into the oncoming lane.
• Areas with major deer problems put up laser sensors along the sides of the road, which detect if an animal crosses the beam and flash lights. A robocar could get data from such sensors to get more advanced warning of animal risks areas.

Getting the deer to move

There might be some options to get the deer to get out of the way. Deer sometimes freeze; a "deer in the headlights." A robocar, however, does not need to have visible headlights! It may have them on for the comfort of the passengers who want to see where they are going and would find it spooky driving in the dark guided by invisible laser light, but those comfort lights can be turned off or dimmed during the deer encounter, something a human driver can't do. This might help the deer to move.

Many people wonder whether robocars will just suffer the curse of regular cars, namely traffic congestion. They are concerned that while robocars might solve many problems of the automobile, in many cities there just isn't room for more roads. Can robocars address the problems of congestion and capacity? What about combined with ITS (Intelligent Transportation Systems) efforts to make roads smarter for human driven cars?

I think the answer is quite positive, for a number of different reasons. I have added a new Robocar essay:

For many years I have had a popular article on what lenses to buy for a Canon DSLR. I shoot with Canon, but much of the advice is universal, so I am translating the article into Nikon.

If you shoot Nikon and are familiar with a variety of lenses for them, I would appreciate your comments. At the start of the article I indicate the main questions I would like people's opinions on, such as moderately priced wide angle lenses, as well as regular zooms.

I've written before about solutions to "range anxiety" -- the barrier to adoption of electric cars which derives from fear that the car will not have enough range and, once out of power, might take a very long time to recharge. It's hard to compete with gasoline's 3 minute fill-up and 300 mile ranges. Earlier I proposed an ability to quickly switch to a rental gasoline car if running out of range.

There's a problem I have seen at a number of free events, particularly "unconference" events which have a limited capacity. There will be a sign-up list, and once it fills up, people are turned away or get on a waiting list. (Some online ticket services now support the idea of free tickets for this purpose.)

I'm at the Pod Car City conference, taking place today and tomorrow in San Jose for PRT developers and customers. Some news tidbits from the conference:

There's news about the world's two small PRT (Personal Rapid Transit) projects, which are both fairly robocar like, in that they involve self-steered rubber tired vehicles.

The AutoNOMOS team at Freie Universität Berlin led by Raul Rojas has shown a demo of their robocar acting as what I have called a whistlecar. Their latest car, named "Made in Germany" performed an autonomous taxi pickup for the press which you can see played out in this video.

This weekend's announcement that Google had logged 140,000 miles of driving in traffic with their prototype robocars got lots of press, but it's not the only news of teams making progress. A team at TU Braunschweig in Germany has their own model which has been driving on ordinary city streets with human oversight. You can watch a video of the car in action though there is a lot of B-roll in that video, so seek ahead to 1:50 and particularly 3:20 for the inside view of the supervisor's hands hovering just over the self-turning steering wheel. There is some information on Stadtpilot here, but we can see many similarities, including the use of the Velodyne 64 line LIDAR on the roof and a typical array of sensors, and more use of detailed maps.

The team at Vislab in Milan has completed most of their Milan to Shanghai autonomous car journey which I have been following. You can read their blog or watch video (sometimes live) of their trip. A lot of the blog has ended up being not about the autonomous challenges, but just the challenges of taking a fleet of very strange looking vehicles in a convoy across Eastern Europe and Asia. For example, they have trucks which can carry their robocars inside, and once decided it was simpler to cross a border into Hungary this way. However, they left driving the vehicles, and the exit officials got very concerned that there was no record of the robocars coming into the country. I presume it wasn't hard to convince them they were not smuggling Hungarian robocars out.

Just released in a New York Times article and sidebar about highways and video, Google has unveiled an internal robot car project that has attained a remarkable level of robotic driving sooner than I and many others had predicted.

In my "New Democracy" topic I am interested in ideas about how technology can change democracy and governance. In California, a rule was passed (curiously needing only a 50% majority) that any ballot propositions that wanted to raise new taxes for specific projects needed a 2/3rds majority to come into effect. I'm in agreement with that. My libertarian bent knows the dangers of letting 51% of the people decide to spend the money of 100% of the people on the flavour-of-the-month.

In this county, a proposition that needs 66% asks for a $29 levy on all properties to pay for medical programs for children. How could anybody vote against that? (I have not examined this proposition in detail, but generally when you see "motherhood" propositions on the ballot, particularly bonds, they have been put there by politicians who have other projects they know would not be popular. So they arrange a ballot proposition to raise money for something nobody could be against, which normally they would have had to spend general revenue on, and this frees up general revenue so they can spend it with less accountability.)

But I digress. And I'm not trying to comment on this particular issue or wishing to come out against medicine for children. But in looking at this proposal, it was clear to me that if 2/3rds of voters wanted it, then you would get the same amount of money if 2/3rds of voters just paid $43.50 (50% more) out of their pockets! No need for a vote (which probably costs quite a bit of money) or asking those who don't agree to pay. In fact, since property owners are probably just a small fraction of the voting population, it might require less than $29 per eligible voter (though not, alas, per ballot casting voter.) With a small amount like this, is there a different way we could do things?

Imagine a contribution system where some sort of publicly funded project could be proposed, with an amount and time period. Each person could register their agreement to pay any amount, including the suggested one, but also less or more. Agreements by registered voters would count as a vote for the plan in addition to being a pledge to pay. (You will see why later.)

(Update: This was written back when Kickstarter and similar companies had just gotten started, you will see similarities to it, and my earlier proposal for congressional voting from 2006.)

The total amount pledged, and the general distribution of it, would be public. People would see if the measure was close to getting its funding target. If it does not reach the target, nobody has to pay. If it reaches the target by a deadline, everybody has to pay what they committed -- in fact it is just added to their tax bill. (This works only with property tax and income tax, not with sales taxes.)

I have put up a page of panoramas from Burning Man 2010. This page includes my largest yet, a 1.2 billion pixel image of the whole of Black Rock City which you will find first on the page. I am particularly proud of it, I hope you find it as amazing as I do.

There was a bit of a stir when Google last week announced that one of the winners of their 10^100 contest would be Shweeb, a pedal-powered monorail from New Zealand that has elements of PRT. Google will invest $1M in Shweeb to help them build a small system, and if it makes any money on the investment, that will go into transportation related charities.

While I had a preference that Google fund a virtual world for developing and racing robocars I have come to love a number of elements about Shweeb, though it's not robocars and the PRT community seems to not think it's PRT. I think it is PRT, in that it's personal, public and, according to the company, relatively rapid through the use of offline stations and non-stop point to point trips. PRT is an idea from the sixties that makes sense but has tried for almost 50 years to get transit planners to believe in it and build it. A micro-PRT has opened as a Heathrow parking shuttle, but in general transit administrators simply aren't early adopters. They don't innovate.

What impresses me about Shweeb is its tremendous simplicity. While it's unlikely to replace our cars or transit systems, it is simple enough that it can actually be built. Once built, it can serve as a testbed for many of PRT's concepts, and go through incremental improvements.

An Italian team has built a prototype robot they call Dustbot which is aimed (in a backwards way) at the deliverbot vision.