cmcmanis@stpeter.Eng.Sun.COM (Chuck McManis) (07/18/90)
In article <9934@pt.cs.cmu.edu> gerry@cive.ri.cmu.edu (Gerry Roston) writes: >In his post, Chuck states that we have a hard time teaching >robots to do things we do easily. This is true. However, >those things which we do well were DESIGNED such that we can >do them well. The same applies to robots. A succesfull design >for a space station, etc, will be one which will allow robots >to assemble/repair it. Fair enough, care to postulate some new designs that are easier for robots. Like, what about fastener technology? Currently robots can weld ok (when the robot base is fixed, we also have to assume a robot in a floating frame of reference) Bolts? Screws? New Adhesives ? Brazing supplies? Let's suppose we have a construction model as follows : Booster A lifts into LEO a "bundle" of building materials. These materials are to be unpacked and assembled by a robot in orbit. The job of the robot is to locate, match orbits with the bundle. Then unpack it and assemble the final component. It must retain and eventually dispose of any packing material. (Tossing it over its shoulder may be viable if this put it into a decaying orbit, plus the robot may wish to use the reaction mass of the packing material to boost its orbit slightly) As a good leveraging experiment analyze the engineering hurdles to having a robot that can approach a "bundle" of materials and assemble it. Can you design it such that it would fit in a shuttle bay? In one of the experimenter pods (The "fly away" specials?) Anyway, can we propose such a paper design? -- --Chuck McManis Sun Microsystems uucp: {anywhere}!sun!cmcmanis BIX: <none> Internet: cmcmanis@Eng.Sun.COM These opinions are my own and no one elses, but you knew that didn't you. "I tell you this parrot is bleeding deceased!"
hollombe@ttidca.TTI.COM (The Polymath) (07/19/90)
In article <139060@sun.Eng.Sun.COM> cmcmanis@stpeter.Eng.Sun.COM (Chuck McManis) writes: }In article <9934@pt.cs.cmu.edu> gerry@cive.ri.cmu.edu (Gerry Roston) writes: }>... those things which we do well were DESIGNED such that we can }>do them well. The same applies to robots. A succesfull design }>for a space station, etc, will be one which will allow robots }>to assemble/repair it. } }Fair enough, care to postulate some new designs that are easier for }robots. ... My current project is to get a robot to emulate a human being operating an automatic teller machine (ATM). Since ATMs aren't built to be used by robots, I've been gaining some useful insights into the problems involved. One of the first things I noticed was that parts of the ATM aren't well designed for people, let alone robots. Not surprisingly, those are the parts that are giving the robot the most trouble. (In particular, inserting envelopes in the depository is turning into a nightmare. People often have to try two or three times to get it right. The robot only gets one try). The one part that was designed specifically to be easy for people (the touchscreen) is the part that gives the robot the least trouble. Conclusion: If we designed things to be easy for _both_ people and robots, both would benefit. As for designing space stations for robot assembly, I'd look at what robots do best and plan accordingly. Robots are good at relatively simple, repetitive tasks. The space structure should be built up from outwardly similar modules that all attach in the same way with the same fasteners. They should also be designed to pack neatly and efficiently so the robot doesn't have to worry as much about bin picking problems. -- The Polymath (aka: Jerry Hollombe, M.A., CDP, aka: hollombe@ttidca.tti.com) Citicorp(+)TTI Illegitimis non 3100 Ocean Park Blvd. (213) 450-9111, x2483 Carborundum Santa Monica, CA 90405 {csun | philabs | psivax}!ttidca!hollombe
ear@wpi.wpi.edu (Eric A Rasmussen) (07/19/90)
In article <18653@ttidca.TTI.COM> hollombe@ttidca.TTI.COM (The Polymath) writes: >As for designing space stations for robot assembly, I'd look at what >robots do best and plan accordingly. Robots are good at relatively >simple, repetitive tasks. The space structure should be built up from >outwardly similar modules that all attach in the same way with the same >fasteners. They should also be designed to pack neatly and efficiently so >the robot doesn't have to worry as much about bin picking problems. Ahh, now that's a simple problem to solve. We should just make the space station out of giant legos. ;-) +---------< Eric A. Rasmussen - Mr. Neat-O (tm) >---------+ +< Email Address >+ | A real engineer never reads the instructions first. | | ear@wpi.wpi.edu | | (They figure out how it works by playing with it.) | | ear%wpi@wpi.edu | +---------------------------------------------------------+ +-----------------+ ((( In Stereo Where Available )))
gerry@cive.ri.cmu.edu (Gerry Roston) (07/19/90)
In article <18653@ttidca.TTI.COM> hollombe@ttidca.TTI.COM (The Polymath) writes: >One of the first things I noticed was that parts of the ATM aren't well >designed for people, let alone robots. Not surprisingly, those are the >parts that are giving the robot the most trouble. > ... >The one part that was designed specifically to be easy for people (the >touchscreen) is the part that gives the robot the least trouble. In the above statement, Jerry SEEMS to be implying that tasks which humans can do easily are tasks which robots can do easily. I do not agree with this. A task which we all face every day is opening a doorequipped with a door knob). For a robot to accomplish this task is quite difficult. It is probably true, however, that there is some subset of tasks which both humans and robots can handle easily. I agree strongly with Jerry's comments about the construction of the space station (although as you are all aware, I am NOT a proponent of said structure!). To strengthen his comments, a study should be done to determine which tasks fall into the catagory of being easy for both humans and robots. Then, given the choice of construction techniques, those tasks which fall into this catagory would be the preferred ones for space applications. -- gerry roston, field robotics center robotics institute, carnegie mellon university pittsburgh, pennsylvania, 15213 (412) 268-6557 gerry@cive.ri.cmu.edu
hollombe@ttidca.TTI.COM (The Polymath) (07/21/90)
In article <9956@pt.cs.cmu.edu> gerry@cive.ri.cmu.edu (Gerry Roston) writes: }In article <18653@ttidca.TTI.COM> hollombe@ttidca.TTI.COM (The Polymath) writes: }> ... }>The one part that was designed specifically to be easy for people (the }>touchscreen) is the part that gives the robot the least trouble. } }In the above statement, Jerry SEEMS to be implying that tasks }which humans can do easily are tasks which robots can do easily. }I do not agree with this. ... And neither do I. Sorry if I gave the wrong impression. As a counter example, one of my robot's tasks is to insert mag stripe cards in the ATM's card readers. Teaching the robot to do that taught _me_ great respect for the human hand and hand/eye coordination. However, if the card readers had been made easier for a robot to use, they'd have been easier for a human to use, too. As for the depository problem, I got that working yesterday. Whoever designed the depository system appears to have been more concerned with keeping things out (rain, soft drinks, dynamite, etc.) than allowing things in. Had it been better designed for use by people, the robot would have had an easier time, too. Had it been designed for use by a robot, people would have had a much easier time with it. }... A task which we all face every day is }opening a doorequipped with a door knob). For a robot to }accomplish this task is quite difficult. ... I'd think that depends on the constraints placed on the robot and the door. There are some amazingly sophisticated pattern recognition systems available off the shelf today. Even if the requirement is "any arbitrary door with a door knob" I can envision a relatively simple system for doing it with mostly off the shelf components (might need a custom gripper or two). It wouldn't be cheap (in a few years it will be), but it would work. -- The Polymath (aka: Jerry Hollombe, M.A., CDP, aka: hollombe@ttidca.tti.com) Citicorp(+)TTI Illegitimis non 3100 Ocean Park Blvd. (213) 450-9111, x2483 Carborundum Santa Monica, CA 90405 {csun | philabs | psivax}!ttidca!hollombe
sandin@uicbert.eecs.uic.edu (07/23/90)
hollombe@ttidca.TTI.COM (The Polymath) writes: >In article <139060@sun.Eng.Sun.COM> cmcmanis@stpeter.Eng.Sun.COM (Chuck McManis) writes: >automatic teller machine (ATM). Since ATMs aren't built to be used by >robots, I've been gaining some useful insights into the problems involved. Ummm.... Why would you want to do this? I mean, obviously, because you can, but really, it seems a little silly... With the proper application of non-robots systems, you can avoid those silly little pieces of paper almost entirely, checks and cash. And besides, what would a robot want to buy? Stephan Meyers c/o Dan Sandin
hollombe@ttidca.TTI.COM (The Polymath) (07/24/90)
In article <1990Jul22.192126.6656@uicbert.eecs.uic.edu> sandin@uicbert.eecs.uic.edu writes: }hollombe@ttidca.TTI.COM (The Polymath) writes: } }>In article <139060@sun.Eng.Sun.COM> cmcmanis@stpeter.Eng.Sun.COM (Chuck McManis) writes: }>automatic teller machine (ATM). Since ATMs aren't built to be used by }>robots, I've been gaining some useful insights into the problems involved. } Ummm.... Why would you want to do this? I mean, obviously, because you }can, but really, it seems a little silly... With the proper application }of non-robots systems, you can avoid those silly little pieces of paper }almost entirely, checks and cash. Note who I work for. TTI writes all the ATM software for Citicorp. That software needs to be thoroughly tested -- preferably in as real world a manner possible. One of my robots spent the last two weeks working the latest release for Citibank, New York. It ran 24 hours a day for tens of thousands of sessions doing things that can't be faked with software (and wouldn't be real world enough if they could). } And besides, what would a robot want to buy? Moly-grease? Upgraded PROMs? (-: Note: I do occasionally find bugs in the controller firmware. The manufacturer (CRS Plus) has been very cooperative in finding and fixing the problems. (Actually, they gave us the last PROM upgrade free of charge). -- The Polymath (aka: Jerry Hollombe, M.A., CDP, aka: hollombe@ttidca.tti.com) Citicorp(+)TTI Illegitimis non 3100 Ocean Park Blvd. (213) 450-9111, x2483 Carborundum Santa Monica, CA 90405 {csun | philabs | psivax}!ttidca!hollombe