Wow. I haven’t posted in a while, mostly because I have been crazy-busy. First, here’s the quick version of the news from last week and earlier: Spain and Canada legalized gay marriage, Supreme Court Justice Sandra Day O’Connor resigned, a branch of Al Qaeda blew up four tube stops in London.

OK. Here’s the main update – since Friday afternoon, I have been in Pittsburgh at the American Association of Artificial Intelligence Conference. We have entered our robot in the scavenger hunt. Ideally, it will go around the floor finding and identifying items. It will follow arrows, find the beach ball, and return it to our table. It will plot everything it finds on a map (we’re only doing Monte Carlo Localization instead of Simultaneous Localization And Mapping, so we had to build the map ourselves Saturday night). However, the lighting conditions are inconsistent at best, and the carpet has a huge pattern on it that hits almost all of our color definitions. We’ll see what happens there.

This place is absolutely incredible. As I remarked to Mac yesterday at the poster session, CMU was giving people free rides on a Segway (for those of you who don’t know, it’s a self-balancing scooter, and it’s awesome), and that was the most boring thing in the room! There is a team from Quebec who has entered a robot in the Challenge event (basically, they need to have the robot attend the conference), and it looks like they’ll actually be able to do it! Their robot, Spartacus, has already waited in line and registered itself for the conference, it has attended several talks, and I believe it can present at its own poster session. It puts our robot to shame, but they’ve had 15 people working for 3 years with a $50,000 budget. Our robot took 4 people $600 over probably 4 months. Still, I’d say we’re better than most of the teams in the scavenger hunt. Nonetheless, Spartacus’ mapping capabilities alone are incredible and awesome, not to mention its ability to understand what you speak to it, read signs on the walls, speak back to you, etc, etc. For those of you who are interested, the way it maps things is your usual idea with a laser rangefinder whose values are compared to the odometry to find out where things are. Ordinarily, this is a bad idea because odometry from your wheels is inherently inaccurate. However, they have an accelerometer in the base of the robot, and they use it to correct the errors in odometry when the chassis turns. They are also using sensor flow data from the rangefinder to correct for the odometric errors when the robot goes forward/backward. It’s incredible.

This morning, Marvin Minsky (who wrote Society of the Mind and has just come out with a new book whose name I don’t remember) gave the keynote speech, and he was surprisingly hilarious. I’ve heard talks about programs that can create debates based on formal argumentation. I’ve seen talks about multi-agent robot swarms learning together (including some guys from the RoboCup soccer league). Stuff about tool use in animals and how we can use what we learn from them to make better robots. Computational models of the cerebral cortex. Markov decision processes. Knowledge manipulation in intelligent systems (this was a really cool talk).

It’s a bit overwhelming, just because there are so many cool things going on here. Tekkotsu had a booth where they were demoing their kinematics on the Aibo (you move its paw around, and using just the sensors in the motors in the joints, it knows where the paw is and can move the head to look at it. This works even when you cover up all the cameras, which blew me away!). There’s a group with a robotic blimp here. The CMU team also has a robot that runs on top of another Segway (it was on Slashdot a couple months ago, but I don’t have time to find links for you right now). This place is absolutely incredible. We have literally been doing stuff at all times here – Friday we checked in and built most of our robot. Saturday, we finished our robot and built a small, demo version with almost no features to put at our table and show off. We then made the map and looked at most of the posters around, and saw some of the other robots. We also started to build our map, which we finished up Sunday afternoon. Sunday evening was the beginning of the scavenger hunt, and we showed off the arrow following and our ability to identify stuff by shape (we can tell the difference between an orange bowl, an orange cone, and an orange dinosaur). We sort of did the beach ball pushing too, but that didn’t work well because our color definitions were off. We left the hotel around 9:30 and found the only restaurant that was still open at 10:00 for dinner. Today, we went to the keynote speech (by Minsky) starting at 8:30, and continued going to talks until 6:30. Dinner at 7:00, then a nap. I think we’ve got the rest of the evening off as well. Whew!

Oh! Another, absolutely fantastic thing we saw today was the Game Playing Competition. 6 groups/individuals have written general game-playing programs! Before each round, the programs are given the rules to the next game (including what a legal move is, how play works, and the object of the game). The programs are then given half an hour to think about how to play, and then they play. I saw them play a variation of Othello on a diamond-shaped board where the winner has the fewest pieces on the board at the end, a round of Chinese Checkers, and a game called Corridor (the object is to get to the other side of the board. During your turn, you can move one square in any direction, or place a wall anywhere you want, presumably to block your opponent). The amazing thing is that these programs could play all of these games well, without having ever seen them before! The winner of the Competition turned out to be a Harvey Mudd alum named Jim who now goes to UCLA. There are actually about 5 HMC alums here, plus the 4 of us, plus Prof Thom (who officiated a set of talks today), and Prof Dodds who is supposed to be here, but hasn’t been seen at the conference since yesterday afternoon. It’s pretty cool.

To sum up, this is the most intellectual stimulation I have ever seen in a single spot, ever. and everyone here is into CS, robotics, and AI! And there are still 2 days left! It’s fantastic. I’m practically in ferret shock.

Leave a Reply

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>