The Robot Diaries

I enjoyed reading the robot diaries which are an account of building some BEAM ^[?] robots from kits. BEAM bots are interesting in that the design sense behind it is much more building from the ground-up (IE, build a robot to resemble an insect) rather then trying to build from the top down. Basically a more evolutionary approach to things.

Comments

BEAM Robots and Battlebots.

[viper21]

I think it would be amusing to create one of these BEAM bots with the sole mission of keeping a 1 meter distance between itself and any other object. Then you put it into Battlebot competition. No one controls it, and all it does is run.

I can see the next breed of Battlebots planning for this escapade. They'll all install forward facing halogen lamps. No BEAM bot can resist the light.

-S

Scott Ruttencutter

I don't believe it.

[paRcat]

I've been building BEAM bots for a few years now, and I was wondering when a BEAM article would get posted here. cool.

For anyone interested in this, especially if you have been discouraged by trying to build other forms of robots, please look into BEAM. While it isn't always as cheap as they like to make it seem, it's much more rewarding than typical CPU-driven robotics.

If you would really like to learn more, you could try Solarbotics and get a kit, or just build your own from scraps. (It's much cheaper that way. :) I have a few that I've built for under $10 US, and none have been more than $30.

But don't be fooled.. even though they are reasonably easy to build, they are worthy of the title "Research Platform". The analog loops that Nv neurons produce can form some strikingly natural paterns. And people tend to like BEAM bots a little more than CPU based versions because they move much more quickly and naturally.

Above all, of course, have fun.

1949 robot technology

[Animats]

The idea of purely reactive robots was first successfully explored by Grey Walters in the late 1940s. He built several mobile, light-seeking, self recharging robots, using relay and vacuum tube technology. These did more than the BEAM robots.

Rod Brooks picked up on this idea, and did some good insect robot work. But then he got hubris, started doing TV interviews, the "Rod Brooks World Tour", T-shirts, and Cog.

When Brooks first gave a talk at Stanford on his plans for Cog, the general idea was to try for human-level AI by building a seated robot body and throwing about 30 MIT PhD theses at the problem. It hasn't worked. I asked Brooks "Why don't you build a robot lizard or mouse?", that being the next step up from the insect work. He said that "he didn't want to go down in history as the guy who built the world's best robot mouse".

This is a classic problem with AI researchers. They get a halfway decent idea, and they start thinking human-level AI is just around the corner. AI goes through one of these enthusiasms every five years or so, some of the main ones having been search, rules, theorem proving, neural nets, and genetic algorithms. All of these are useful, and all have hit a ceiling beyond which further work doesn't produce much improvement.

I tell people we're probably going to have to claw our way up the evolutionary ladder, and the next step is the lizard brain level of intelligence. This is happening, amusingly, in the game world, where opponent control AI has to solve the basic problems of life: not falling down, not bumping into stuff, and back-seat driving the machinery that controls those tasks into getting something done.

I would have been more impressed...

[cr0sh]

Had the writer of the diary actually built the machines...

Instead, she intones that her boyfriend and her roommate ended up putting them together, and the roommate was the one actually soldering - what, was the boyfriend there for moral support (apparently so, she says later on one goes home after the racer did not immediately work, which could only mean the boyfriend, since the roommate is at home)?

Instead, she says she served drinks. Then, the rest of the article goes into really inane observations of the machines "at play". The only fun observation was of the photovore avoiding a grape, and getting to the lamp.

All this individual ended up doing was writing an article. I would have been more impressed had she tackled the soldering iron, made a few mistakes, learned how to solder, and build the racer. Even if it didn't work right, it at least would have shown that she tried to learn something completely new - instead of passively letting life go by.

Furthermore, she doesn't feel these devices are really worth the effort put into the building of them. You can tell by the tone of the writing. She talks about setting up an environment filled with various knick knacks and things. Why doesn't she get it about BEAM - hobble the damn thing! Put tape on one of the photovore's motors, see how it works around this "impediment"! Geez, is experimentation that difficult?!

People, if you want to know more about BEAM, and want to play with it yourself, avoid this article. Here are a few links to check instead:

http://www.solarbotics.com/ (she could've at least provided this link!)
http://www.nis.lanl.gov/projects/robot/
h ttp://www.geocities.com/frankendaddy/BEAM.html

Worldcom - Generation Duh!