Slashdot Mirror
Denver Couple Unveils Homemade Service Robot
An anonymous reader writes "Jim & Louise Gunderson, owners of a Denver-based computer software tool development company, have finally unveiled their autonomous robot, Basil. Basil is completely home built, runs Linux with some instructions in Java, uses a sonar-based 'reification' logic system, and can go get you a beer or a pot of tea. Quoting: 'The plan is this: The Gundersons will ask Basil to go to the bar, request a couple of stouts from the bartender, and then, once they're placed on the titanium tray perched on his head, bring them back to his creators. They haven't told him how to do this — there's no set script in his processors that tells him to roll a certain distance southwest, speak a certain command, then come back. He'll have to figure it all out on his own, using a basic knowledge of bars and beers and so on, reasoning skills and an ability to understand certain parts of the world. When his sonars capture the image of a person, for example, he knows it's a person, not just a nameless object to be avoided. And he knows that, in this case, that person wants a beer.'"
"I recognize a person 69cm away"
"I recognize a wooden chair"
Right. Using sonar, the robot is able to determine the composition of the chair.
Given that the robot's speech patterns are not broken at all, and that it speaks in complete sentences, it seems more likely that this is a blinkenlites contraption with a very human person controlling it the whole time.
Don't let that bowtie fool you. I know a Dalek when I see one.
Beer? Great! But what is beer without titties?
http://www.youtube.com/watch?v=lwb1s1DYnDU
Where's the Kaboom?
There's supposed to be an Earth-shattering Kaboom.
Any UBC EEs from 83/84 will remember our robot, also called Basil, because it was somewhat faulty (Fawlty).
"...runs Linux with some instructions in Java..."
Uh-oh, they used the J-word. Wait until the Slashdot Religious Order gets their hands on them.
It's simple: I demand prosecution for torture.
welcome my beer. Thanks, robot underling.
Tic-Tac-Toe, Global Thermonuclear War, and relationships all have the same winning move.
It looks like the sensors are dumb ranging sonars at four heights. Those are very crude sensors; all you get is the range of the nearest solid object in a 30 degree cone. You could probably separate walls, tables, chairs, and humans with that, at least some of the time. It won't ever work very well. People have been fooling with those things since the 1980s. (The usual sonar sensors are left over from Polaroid auto-focus cameras. Very few robotics people have tried to do serious sonar processing, like submarines or bats.) You're just too information-starved. Vision, though...
There's been much more progress in the last five years than most people realize, though. SLAM works now. Vision algorithms actually work. Low-cost inertial devices work. We're starting to see the payoff from the DARPA Grand Challenge, which gave robotics a serious and needed butt-kick.
He'll have to figure it all out on his own, using a basic knowledge of bars and beers and so on, reasoning skills and an ability to understand certain parts of the world.
This strategy seemed to work very well for George W. Bush.
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
Unless it can swim to Europe Hows it going to obey a command to fetch a REAL beer?
At Denning we had a mobile robot security guard. It could roam a factory or warehouse looking for intruders. it had sonar, radar, and other things.
Notifying people of appointments, delivering small objects, and serving drinks is not only possible, it is probably the easiest set of tasks that you can do.
I have a project on-line that allows you to build a basic robot for $500. It has PWM motor control and basic tips on building the base. It uses a PS/2 mouse to do wheel encoders. (cheap) and using a USB A-D/D-A board to control stuff. (I won't give the URL for fear of slashdotting my server.)
So, my two points: 1) It is possible they are doing what they say they can do. 2) Its fairly trivial if you have the time to waste.
I for one, really like the way they decided to proceed when making this robot. It works by a healthy mix of abstracting and trial and error.
Let's take the wooden chair, that is used as an example in TFA. As far as I understand it, learning about it and using this information for the robot goes like this.
They put the robot in front of the chair and let it use it's sonars on it from different angles and distances. I imagine that in the case of a typical wooden chair with a back it sees four points for the legs and a line for the back. At least I believe that it abstracts it as such. For the first time it will be input to it that the thing it sees is a wooden chair and it knows that all things that have four points about so far from each other in a squared manner and have a line above two of the side points can be regarded as a wooden chair. If it sees another chair made of metal without the back for example, it might consider that to be a wooden chair as well because it's similar enough and in that case the makers correct it's assumption and say it's a metal chair. Sure, it will start to think that all the chairs without the back are metal chairs, but if that's the case in their home, so what, it's right. If it understands anything wrong enough that it fails at its task it can always be corrected and its knowledge about the world as it sees it will increase. Now when performing tasks it can treat the chair as an abstract object, now that it can recognize it. It can memorize where it stands, it can learn to avoid it or push it or whatever, as long as humans correct its assumptions and choices. Now these abstractions could be abstracted even further. The idea is to let it do very simple things and then combine them into larger tasks, much like programmers think about and solve programming problems: If you want to solve a large problem and you don't know how to, you break it into smaller pieces until you get a piece, that is simple enough to be solved. You solve it and see the next piece. Then you combine the solutions to a solution to the bigger problem and you finally end up with the first and biggest problem getting solved. This robot 'learns' the exact opposite way.
It seems to me that the biggest concern in this case is abstracting the objects it 'sees' into such a form, that they take minimal memory but can still be used in the recognition process.
That came out as ranting. I have no knowledge in the subject and have no idea what I'm talking about but that should make this a good enough Slashdot comment.
I hope it isn't fawlty...
Go east, to a place called Klamath. K-l-a-m-a-t-h. Find Vic. V-i-c. Ask for beer. B-e-e-r. *sigh* You are the chosen one. Find the beer. Be our salvation.
That's a simple algorithm:
if (object == person)
wants_beer = 1;
Sure there is going to be some margin of error in that algorithm, but it's going to be right most of the time.
Modern military-grade sonar can EASILY tell materials just by the sound quality bounceback. So can whales, dolphins, bats and pretty much any creature with ears, including humans.
Try this: Walk into an empty room with sheetrock walls and a wood floor and clap your hands. Now do it in a similar room with a tile floor and wood paneling on the walls. Now an all-concrete cinderblock room. You will notice that, even though the source sound is the same (your hands clapping) the return sound has a different quality in each room. However, the sound quality in a single room will always be the same, regardless of the number of claps you make.
Now try it blindfolded, and see if you can differentiate the rooms. You will be able to, unless you are hearing-impaired in some manner.
It's the same principle with the robot. Once taught about an item, it can continue to identify the item even though it can't "see" it.
Official Heretic from the "Church of Global Warming". Proven right thanks to whistle blowers. AGW = Flat Earth Theory
Wouldn't it be more like "sudo get me a beer"?
However, when the robot is given a goal ("deliver tea to the conference-table-area") the mental model is used to generate a symbolic representation of the world-as-it-is, along with the representation of the world-as-it-is-desired. At this point, the tag "wooden-chair" is used to extract information from the semantic memory (an ontology of facts and behaviors) and the linkages in the ontology allow Basil to reason about chairs with respect to the current goal. So he knows that chairs are generally stationary - they stay put, as opposed to people who move on their own; he knows that wheeled-chairs can be pushed out of the way, but the wooden-chairs can't, and that people can be asked to move, but chairs can't.
So at this point the label begins to be less arbitrary since it is now embedded in a complex knowledge structure. If we gave the chair the label 'battleship' (and if we had information about battleships in the ontology), Basil would generate different behaviors with respect to the object.
The classification scheme is fairly simple - primarily because his sensor modalities are thin. He builds a set of representations (classic pattern based templates) and uses these for both recognition and preafference (projecting what the world should look like).
When he is learning a new object, he checks to see if the patterns are mutually exclusive or if two or more objects can be classified from the same sensor data. If there is no way to distinguish between two classes, he reports both as possibles. These get loaded into the mental model and as he gets more views of the object the winnows down the possibilities.
So he is using multiple time and space separated views and 'thing constancy' as a principle to help him classify. There is a whole lot more detail in our book.
Basil is designed to learn from his experiences. He maintains a complete episodic memory at present. The task for the next year is to enable him to analyze these memories and generate new sensor representations, to subdivide existing representations, and to add new facts to his semantic memory. The tools that we will use will be a mix of standard machine learning techniques along with a technique that Louise developed for environments where not all features are salient to the classification.
Jim Jim