weintrb@cheops.cis.ohio-state.edu (Mike Weintraub) (01/05/90)
I am looking for information/experiences/preferences people have with commercially available robots. The more sophisticated it is, the better. In fact, if anyone has one they would like to donate (preferably something like Heath's Hero with nice programmability features like an rs-232 port connection to the transmitter), that would be nice. Thanks. -- Michael Weintraub Laboratory for AI Research The Ohio State University weintrb@cis.ohio-state.edu
Nagle@cup.portal.com (John - Nagle) (01/08/90)
Personally, I've done a blocks world with an IBM RS-1 robot, and would recommend against using one of those old monsters, even if you can get it free. Dealing with the IBM Series I and the AML interpreter is too restrictive, the system only executes about 30 interpretive statements per second (!), and the robot itself is overly complicated for a research environment. Also, the hydraulic pump is noisy and requires chilled water for cooling. The Adept Technology machines, with their built-in vision systems, seem to be the machine of choice for routine manipulation work today. These are SCARA geometry machines. They run about $60K, and are controlled by a Mac II. There's one at the Center for Design Research at Stanford, and the people who use it seem to be happy with it. Contact Prof. Mark Cutkosky at Stanford for details. For mobile robot work, check with Rod Brooks at MIT and find out how he's doing with those little battery-powered platforms that outfit near Boston is selling him. But it's just a raw platform, without electronics. Cybermation (Roanoke, VA) has a line of largish mobile robots, radio communications gear, and such. They appear to build solid iron. They have a narrow-beam ultrasonic system in the 1MHz range that is rather impressive, although you only get one range reading in a narrow cone per cycle, so it's a slow way to map the environment. The Hero 2000 is not all that useful. The sensors are too limited and the navigation system is too weak. But somebody in Bernado Hubermann's group at PARC had one hooked up to a Symbolics around 1987. Don't know whatever became of that effort. Good luck. This is a very difficult area to work in. John Nagle