behara@turing.cs.rpi.edu (Albert Behar) (10/29/90)
I'm working to build a mobile robot research platform and have thus far used dead reckoning (optical encoders on wheels) to determine the robots global position. The results have been good but the errors increase as time goes on. Now, I wish to increase the position tracking accuracy because the platform will soon need to support a vision platform. Does any one have any idea where I can get some sort of tracking device or how others have treated this problem? Thanks in advance, please send any responses directly to me. ++++++++++++++++---------------------------------+++++++++++++++++ Albert Behar email:behara@turing.cs.rpi.edu NASA Center for Intelligent Robotic Systems for Space Exploration Rensselaer Polytechnic Institute, Troy, NY 12180
roy@phri.nyu.edu (Roy Smith) (10/29/90)
behara@turing.cs.rpi.edu (Albert Behar) writes: > Does any one have any idea where I can get some sort of > tracking device or how others have treated this problem? How much money do you want to spend, and how accurate/precise do you need it? One possibility might be to buy a marine LORAN receiver. You can get them for as little as about $400 now, complete with some sort of RS-232-ish interface, runs off of 12V, not much larger or heavier than a big calculator. Absolure accuracy isn't very good, but precision and repeatability is very good, down to perhaps 100 feet or so. If you want to do better than that, and have really deep pockets, you might think about a GPS receiver. There are currently only a few on the market, with price tags an order of magnitude greater than the LORAN sets, but they will get cheaper as time goes on (it's a new technology) and are supposed to be accurate down to like 10 feet, or something like that. I suspect that neither are really what you are looking for, but I figured it might give you some ideas. /roy -- Roy Smith, Public Health Research Institute 455 First Avenue, New York, NY 10016 roy@alanine.phri.nyu.edu -OR- {att,cmcl2,rutgers,hombre}!phri!roy "Arcane? Did you say arcane? It wouldn't be Unix if it wasn't arcane!"
gerry@frc2.frc.ri.cmu.edu (Gerry Roston) (10/29/90)
The two solutions proposed so far are probably the best: loran and gps. if this is not what you want, here is a cheap and sleazy method: Get 3 yo-yo pots (more info available if requested). these are precision potentiometers/tachometers hooked up to a spool if string (similar to a key-chain ...). Anyway, locate three of tehm in the world, attach the strings from all three to your robot, and you now know its position. Well, I did say sleazy. gerry -- gerry roston, field robotics center robotics institute, carnegie mellon university pittsburgh, pennsylvania, 15213 (412) 268-6557 gerry@cs.cmu.edu
bobc@hplsla.HP.COM (Bob Cutler) (10/30/90)
If you are operating in a fixed environment you might consider the following methods. These are methods which are used in an experimental platform based on a mobile robot build by a company called Denning. Last I heard, the robot(s) lived in the EE department at the University of Washington. 1. Ultrasonic sensors which can be used to identify known objects such as walls, doors, etc. These sensors are easily fooled by reflections from hard surfaces. For instance, to ultrasonics, a hardwood door looks like a mirror. The Denning Mobile robot has 24 Polaroid sensors. 2. Lasers can be used to scan walls for reflective tape. Using several pieces of tape, you can create bar-codes which can be read by the laser. The angular width of the scan can be used to get a rough estimate of distance. 3. Beacons. These work kind of like a lighthouse. Each beacon hangs on a wall and emits a unique code. These can be implemented using LED's. The robot would determine its position by locating and identifying two or more beacons. Since the absolute position of each beacon is known, the position of the robot can be determined from the angular positions of the beacons relative to the robot.
hollombe@ttidca.TTI.COM (The Polymath) (10/31/90)
In article <$M7%L=#@rpi.edu> behara@turing.cs.rpi.edu (Albert Behar) writes: }I'm working to build a mobile robot research platform and }have thus far used dead reckoning (optical encoders on wheels) }to determine the robots global position. The results have been }good but the errors increase as time goes on. } }Now, I wish to increase the position tracking accuracy because }the platform will soon need to support a vision platform. Since you're going to have a vision platform, why not use it for actual ded-reckoning [sic]? (Pedantic Note: That's not a typo. The "ded" is short for "deductive." Ded-reckoning is navigation by reference to known landmarks, or so I was taught in ground school). There are some spectacular pattern recognition systems available off the shelf these days. Why not have your robot look around to see where it is? -- The Polymath (aka: Jerry Hollombe, M.A., CDP, aka: hollombe@ttidca.tti.com) Head Robot Wrangler at Citicorp(+)TTI Illegitimis non 3100 Ocean Park Blvd. (213) 450-9111, x2483 Carborundum Santa Monica, CA 90405 {csun | philabs | psivax}!ttidca!hollombe
cgordon@vpnet.chi.il.us (Crash Gordon) (10/31/90)
>Author: [Albert Behar] >Does any one have any idea where I can get some sort of >tracking device or how others have treated this problem? How large is the area to be covered by this vehicle? If it's room-sized, you might look into the Power-glove method; either sticking with acoustics, or moving into RF for larger areas. This won't give you global coverage (unless you've got some _SERIOUS_ acoustic transducers ;-) but it might get the job done for, say, a warehouse application. Other small-scale solutions involve buried wires (linear or grid pattern) and painting of distinctive reference marks on the vehicle path. You might also look into a doppler based DR system for better accuracy, periodically updated by LORAN/GPS to handle long-term error. If you want total independence, you're probably breaking new ground. NASA has a couple of self-guided vehicles, but I think top speed is around 3 mph, cost is to standard gov't specs, and the brains are networked Sun workstations... ----------------------------------------------------- Gordon S. Hlavenka cgordon@vpnet.chi.il.us Disclaimer: Yeah, I said it. So what?