Stiquito: A Small Nitinol-Propelled Hexapod Robot The Computer Science Department's Robotics Laboratory at Indiana University (Bloomington) has designed, built, and tested "Stiquito," a six-legged robot that you customize by adding sensors, controllers, power sources, etc. The robot provides an inexpensive platform to study computational sensors, subsumption architectures, neural gait control, emergent cooperative behavior, and machine vision. It is currently being used for research at IU, and, at a ratio of one robot per student, in "VLSI for Robotics" and "Machine Learning" classes. Stiquito is small (3cm H x 7cm W x 6cm L) and simple (32 parts) because its legs are propelled by nitinol actuator wires. Each leg has one degree of freedom. The robot walks up to 10 centimeters per minute and can carry a 9-volt cell, a MOSIS "tiny chip" and power transistors to drive the nitinol actuator wires. Alternatively, power and control can be supplied through a tether. A single robot costs between $10 and $30. The pro-rated cost of materials to construct robots in volume (200 or more) is less than $3. The design can be replicated to build complex arthropods, or colonies containing hundreds of insectoid robots. IUCS Technical Report 363a, "Stiquito: A Small, Simple, Inexpensive Hexapod Robot -- Part 1. Locomotion and Hard-Wired Control" contains detailed instructions to build the robot and a hard-wired, manually-operated controller. The instructions are illustrated with 80 life-sized figures that include dimensions and enlarged details. Materials, tools, and vendors are listed. Special sections include construction skills for novice users, how to fix mistakes, and troubleshooting problems found in real-world experience with the robot. Stiquito and its manual controller are assembled by crimping and press-fitting parts. No soldering is necessary. The resulting robot is very reliable. A tethered Stiquito walked for over 100 hours. During the test each leg made 300,000+ 2-3mm strokes. The leg movement was reduced because power was supplied during the test by a 6v AC/DC adapter. With a 6v "J" battery typical leg movement is 3-7mm. Battery-powered leg movement was unchanged after the test. Anyone may build and use Stiquitos in any quantity for educational or research purposes, but Indiana University reserves all rights to commercial applications. Part 1 of the technical report is now available and can be obtained in two ways: 1. ELECTRONICALLY ----------------- Use anonymous ftp to obtain the technical report free. IU's ftp site is: cs.indiana.edu Log in as "anonymous" with username@host or "guest" as a password. The report will be in one or the other of two directories: pub/ pub/techreports The report is available in two formats: stiquito.ps.a{a,b,c}.Z (PostScript) stiquito.w4.a{a,b,c}.Z (Macbinary Word 4 for IBM PC users) stiquito.hqx (stuffed, binhexed Microsoft Word 4 for Mac) Unfortunately, some of the illustrations in the PostScript file print with minor errors (patterns missing, and some polygons deformed), but they are still understandable. Some users have trouble printing large postscript files. These files are 1.2 Mb+. Some postscript printers will not print the files, even though the Macintosh library routines are prepended on the .ps. file. E-mail for a FREE copy of the tech report if you can't get the file to print. 2. BY MAIL (with options for a kit and a videotape) --------------------------------------------------- Send your request for the report with payment to: Computer Science Department 215 Lindley Hall Indiana University Bloomington, Indiana, 47405 To receive the technical report only send $5 US PRE-PAID and add the following line to the address: ATTN: TR 363a To receive the technical report and a kit containing all materials needed to construct Stiquito and its manual controller send $15 US PRE-PAID. Add the following line to the address: ATTN: TR 363a Stiquito Kit The kit alone is available for $10 US PRE-PAID. Add the following line to the address: ATTN: Stiquito Kit The charges cover materials, a kitting fee, the technical report, and mailing. The kit is provided as a non-profit service by Indiana University. The author receives no fee or reimbursement from the sale of the kit. The CS department has also videotaped the assembly of a Stiquito (including the author making some mildly embarrassing mistakes during construction! :-). It is available separately for an additional $10 US PRE-PAID. Just add VIDEO to the attention line. For example, if you purchased the TR, kit, and video then $25 US would be enclosed and the following line would be added to the address: ATTN: TR 363a Stiquito Kit VIDEO The kit has been modified to include MORE materials at NO extra cost to be more forgiving if you make a mistake. A benefit if you DON'T make a mistake is that two kits will produce THREE robots, leaving extra 9v cell terminals. Part 2 of the technical report, "Sensors and Controllers," is in preparation. It describes analog and digital controllers, legged Braitenburg vehicles, and sensor and communicatons interfaces to Stiquito. Part 3 of the technical report, "Experiments with Arthrobots," will describe a variety of arthropod robots and report on experiments in emergent cooperative behavior using a colony of Stiquitos. Questions about Stiquito should be sent to jwmills@cs.indiana.edu. There are many different ways to design smaller, larger, or multiple-degree-of-freedom nitinol-propelled robots, legs, tentacles, grippers, etc. Some we've tried, most we haven't. We hope you will be encouraged to use Stiquito as a stepping-stone to better and more effective miniaturized robots. ------------------------------------------------------------------ This research was funded in part by NSF Grant MIP-9010878 "Lukasiewicz Logic Arrays" awarded to J.W. Mills, and NSF Grant CDA-8852304 "Incubator Laboratory Development" awarded to D. Gannon, G. Rawlins, and J. Smith. NSF's support is gratefully acknowledged. Standard disclaimer for opinions, etc. expressed in this posting. ------------------------------------------------------------------ - J. W. Mills "Sic ago!" In article <1992Sep24.113622.24133@news.cs.indiana.edu> I wrote: > > Send your request for the report with payment to: > > Computer Science Department > 215 Lindley Hall > Indiana University > Bloomington, Indiana, 47405 > > [various options & prices deleted] Please make checks or money orders out to "Indiana University." It is not a good idea to send cash, although some persons have explained they have no other choice. Please include payment with your request. The university isn't set up to bill through credit cards for small projects like this, nor can purchase orders be processed easily or in a timely fashion. If you want to use a purchase order, the best way is to have your organization's purchasing department write a check for you to send. Don't worry about currency fluctuations, we'll send your kit if the amount is close. (After all, the money markets change every few minutes, and the point is to get a robot to you as quickly as possible.) Everything is shipped First Class in the US, and Airmail elsewhere, so the TRs & kits should arrive reasonably quickly. Federal Express is possible if you pay for it, but costs as much as or more than the robot. I'd advise against it. If you need a robot quickly, ftp the TR, print page 3 (the materials list), and head on down to your local hobby shop. I'd like to repeat my apology for the problem about the PostScript file. It was (and is) huge -- over 1.2 Mb --, so I split it and compressed it. The files are in /pub as: stiquito.ps.aa.Z (uncompress & cat to recover) stiquito.ps.ab.Z stiquito.ps.ac.Z The PostScript files now have the Mac PS libraries prepended to them so anyone can print them on (hopefully) any PostScript printer. A UNIX-compressed MacBinary Word 4 version is also available for those running Word on IBM PCs/clones: stiquito.w4.aa.Z (uncompress, cat, and uudecode to recover) stiquito.w4.ab.Z stiquito.w4.ac.Z Be sure to set ftp to binary before retrieving these files! The stuffed & binhexed Word 4 document is still available as: stiquito.hqx which is a _text_ file, not a binary. Please send me e-mail if you have questions that aren't answered, and I'll respond as quickly as I can. I'm looking forward to seeing clever and innovative designs for miniature robots. If there's enough interest, maybe even a nitinol-fueled Stiquito dragster race at Mark Tilden's BEAM robot olympics. Let's see, how to build a nitinol-propelled robot grasshopper... ------------------------------------------------------------------ - J. W. Mills "Sic ago!" From: nagle@netcom.com (John Nagle) Subject: Re: Stiquito, A Small Nitinol-Propelled Robot The document describes a rather simple version of the thing. The "robot kit" actually yields a six-legged device with the three legs on each side tied together electrically. The legs advance in a rachet-like fashion, under the control of a pair of manually-operated pushbuttons. References are made to future VLSI-controlled versions, but that's not what you get. A $100 kit with more smarts would be much more interesting. John Nagle In article nagle@netcom.com (John Nagle) writes: > The document describes a rather simple version of the thing. John is correct. Stiquito _is_ very simple, but was intended to be simple. This is because we planned to, and are, using the robot in volume ("ant colony"). Effort was made to make the walking part buildable by students who might not have much mechanical expertise. A consequence is that this "robot" will not compare favorably to those with multiple-degree-of-freedom legs, or sophisticated grippers, etc. On the other hand, Stiquito is a good starting point to use to investigate nitinol-actuated devices. A more complex predecessor, "Sticky," has legs with two degrees of freedom, nitinol actuators inside the body tubing, and can carry 6 to 8 ounces. It is correspondingly more complex, and gains structural stability with modularity at the expense of complexity. Although Sticky can be assembled by hand, and disassembled with a bent paper clip, it resembles a Chinese puzzle. I'll be glad to send photographs of Sticky, or perhaps digitized photos of a disassembled robot, with the assembly sequence, that could be used to show how to make one. The cost of Sticky, minus electronics, is about $20. Sticky + miniboard may be more what the robotics community currently expects. I'll help all I can to satisfy that expectation. >The "robot kit" actually yields a six-legged device with the three legs >on each side tied together electrically. The legs advance in a rachet-like >fashion, under the control of a pair of manually-operated pushbuttons. Again, John is correct. However, the legs do not have to be hard-wired in a tripod gait. As is, the kit can be used to experiment with other gaits. The manual controller is also useful to verify that the legs work. And, sad experience with early oscillator circuits that powered the legs for too long, damaging them, led me to build the manual controller so I could check the robot out quickly. >References are made to future VLSI-controlled versions, but that's not >what you get. This is true. The technical report, TR 363a, is one of a series of reports on work with these robots. Please accept my very sincere apologies if you feel you have been mislead. I pointed out in the original post that you had to add a controller, etc. I understand the disappointment that Stiquito doesn't do more than it does, but I have also received very encouraging notes from others who plan to use Stiquito-clones in their research. Don't give up yet because... A mailing list is being set up by folks at UC Berkeley to monitor advances in projects, designs, and answer FAQs about Stiquito-style robots. Subscribe to this list (instructions should be posted soon), and you'll get all the news as it breaks. I plan to post descriptions of the VLSI circuits that John mentioned within a few months in a draft of TR 363b (please note the "b"), "Sensors and Controllers." Other circuits will be posted next week. Already built and tested is a Braitenburg Bug based on Stiquito that is phototropic, and programmable as either a photophil or a photophobe. It uses two independent oscillators driven by photo- transistors. The photo-transistors can be placed in the circuit to make it stroke-length-variable, or stroke-rate-variable. The circuits are similar to the incandescent lamp flasher found on page 17 of Forrest M. Mims III pamphlet "Engineer's Mini-Notebook Optoelectronics Circuits" available from Radio Shack as catalog number 276-5012. Although nitinol is similar to the tungsten filament on a lamp, there are three legs to consider: add extra 2N2222's, one for each leg, to avoid overheating. Be sure to hook up the legs before powering the circuit or you risk damaging the transistors. Be very careful that the circuit oscillates slowly, or you may damage the legs. Almost finished and next to be tested is an IBM PC interface that can drive Stiquito from a serial port. It may be possible to drive Stiquito, each leg independently, from a PC parallel port. I haven't tried it yet, so check the load imposed on the port as a whole if all six legs are driven at once: that will draw about 1.2 amps, more than the port may source. The VLSI circuits John mentioned are the topic of current research. The first types are field-programmable gate arrays from Actel that my students and I are using to implement frequency-modulated controllers for the robot. The idea is to sense light and pressure then use the inputs to generate trains of pulses at intervals to cause the legs to walk at different rates, turning the robot, moving it backwards, etc. The legs have ratchet feet angled in different directions so the robot can walk backwards, sideways, etc. The second type of VLSI circuits are custom analog circuits fabricated through MOSIS. They are called "Lukasiewicz Logic Arrays," and have been the topic of my research since 1990. I built the robots as a platform to work with these circuits. These "LLAs" use a continuous-valued logic that can be interpreted as both a logic and an algebra. Using the algebraic interpretation, LLAs act as single- transistor directed differentiators. Using the logical mode, the LLAs can be used to build fuzzy inference engines. Both modes use the same device structure. In my paper that appeared in the 1992 22nd Symposium on Multiple-Valued Logic ("Area-Efficient Implication Circuits for Very Dense Lukasiewicz Logic Arrays), a simple but fast silicon retina was briefly described that used one transistor per pixel as the computational element. The rest of the computation was done with wire. Using LLAs, and a fuzzy controller I simulated this summer, my next step is to put both on a MOSIS tiny chip, and use it to drive Stiquito as a Braitenburg Bug that responds to more complex patterns, such as textures. But until the thing walks and follows a texture, it's vaporware. I'd feel comfortable sending you published papers, but contact me directly for reports on the LLA project. > > A $100 kit with more smarts would be much more interesting. > > John Nagle Very, very true. For now, the ftp access to the report is working. Try building a Stiquito, (the sources of all materials, including telephone numbers are listed in the technical report -- you DON'T have to buy the kit by any means), subscribe to the mailing list, and let your creativity flow. And have fun, everybody here who has seen Stiquito the Braitenburg Bug walk has immediately thought of six impossible things to do before breakfast. You can ftp in directly to the net address of cs.indiana.edu: 129.79.254.191 There have been some connections broken due to timeouts. I've split and compressed the PostScript & raw Word 4 files, and will do the same for the .hqx version if necessary. If you have problems with timeouts, our net guru, Bruce Sheu, has kindly offered to help. Send him mail at shei@cs.indiana.edu. To get the tech reports, connect using anonymous ftp. The reports are in /pub. Set the ftp mode to binary to access the PostScript and raw Word 4 files: stiquito.ps.aa.Z (restore with uncompress, cat) stiquito.ps.ab.Z stiquito.ps.ac.Z stiquito.w4.aa.Z (restore with uncompress, cat, uudecode stiquito.w4.ab.Z to get a MacBinary Word 4 document) stiquito.w4.ac.Z Here are answers to some questions on the report: Does the robot walk autonomously using materials provided in the kit? -No, but it is fairly easy to attach a small printed circuit board to the aluminum tubes that extend from the robot's back. A draft of Part 2 containing two transistorized oscillators controlled by ECG 3034 phototransistors will be posted sooner than I'd expected, due to these requests. These circuits, and a 9-volt cell, turn Stiquito into a legged Braitenburg photophilic or photophobic bug. What is nitinol wire? -Nitinol, sold under various tradenames (BioMetal, Flexinol) is an almost-50-50 alloy of nickel and titanium that changes between two crystalline structures as it is heated and cooled. It can exert useful force during the transitions, and make as many as 5,000,000 transitions before it ceases to work (maybe more). Passing a current through nitinol generates enough heat to cause it to contract. A spring or other force is necessary for the wire to expand back to its original longer length when it cools. Nitinol can be manufactured in many shapes (eyeglasses, braces for teeth, etc.), but the cost of special-forming is said to be high, how high I don't know. The wire costs $10 per meter from Dynalloy, Inc., Mondo-Tronics, & a few other manufacturers, but drops to $5.50 in 101 meter lengths. The September-October issue of "Midnight Engineering" has a cover photograph of a similar nitinol-propelled robot developed independently by Roger Gilbertson of Mondo-Tronics. I think that Stiquito looks a little sleeker, but then I'm predjudiced! In article nagle@netcom.com (John Nagle) writes: References are made to future VLSI-controlled versions, but that's not what you get. furthermore, with a 50 GRAM payload, it is going to be hard to put enough batteries on the beast to take more than a few steps. tethered operations are more interesting. people should also note the 10cm per minute walking speed. The document describes a rather simple version of the thing. The "robot kit" actually yields a six-legged device with the three legs on each side tied together electrically. The legs advance in a rachet-like fashion, under the control of a pair of manually-operated pushbuttons. References are made to future VLSI-controlled versions, but that's not what you get. I too was somewhat disappointed by the simplicity of the "robot" described, even though I had already noticed the 20 cm/minute walking speed. Further consideration tempered by reaction. It's a neat device. Furthermore, most "biometal" kits selling for about $30 and contain an inch or two of nitinol wire. The stiquito kit, with about 16 inches of wire, is a "good deal". (eg you get what you pay for, but that's not so bad...) Stiquito (as described) is a clever mechanical engineering hack, with interesting posibilities. A $100 kit with more smarts would be much more interesting. To some. I think that Mr Mills, and Indiana U, ought to contact that toy company with all the neat "robot" kits. There's money to be made. I am curious about a couple of points: 1) My Flexinol(tm) price list says the .004 inch wire used by stiquito is "Special order". Was it just left over from "Sticky", or is there more behind the choice? 2) Since the instructions imply that a 3V power source can overheat the wire, using a 9V battery for the demo controller seems perilous indeed. > 1) My Flexinol(tm) price list says the .004 inch wire used > by stiquito is "Special order". Was it just left over from > "Sticky", or is there more behind the choice? Dynalloy and Mondo-Tronics sell evaluation kits containing Flexinol(tm) in a variety of diameters. Of the wires tested by building legs, the .004" diameter wire drove the leg reliably, even when the leg was held with pliers, without overstressing the wire. The wire consumed between 180-220 mA of current, acceptable for battery operation. It was also easier (for me) to knot the .004" wire than the smaller wire, and the knots were tighter than with the .006" wire. I have not tried doubling smaller wires, nor have I tried the .003" wire. Both alternatives should work, though. Reporting on battery life for Stiquito, a fresh 6-volt Eveready alkaline "J" battery drove the robot and the two oscillator circuits continuously for 2 hours 25 minutes. The robot walked in a circle about 2 feet in diameter, but I didn't watch it the whole time so I can't tell how far it went. The walking rate did decrease, but the sun was beginning to set, too, by 5:30, leaving the test site in more shade than at the start. While I did watch it, the robot moved a measured 45cm in 5 minutes for a rate of 9cm/minute. > > 2) Since the instructions imply that a 3V power source can > overheat the wire, using a 9V battery for the demo controller > seems perilous indeed. > The report warns about this. Single actuator wires should be tested with 3v, but the hard-wired tripod, with three wires, can be operated by _briefly_ closing the switch on the manual controller with a 9v cell attached. Don't continue to hold the switch down if nothing appears to happen, because if the actuators are connected but slack, they will overheat and be damaged. The same thing may happen, but is less likely, with a single wire and two fresh AA cells. Article 3405 of comp.robotics: Newsgroups: comp.robotics Path: news.cs.indiana.edu!sgiblab!swrinde!zaphod.mps.ohio-state.edu!cs.utexas.edu!convex!constellation!a.cs.okstate.edu!rjs From: rjs@a.cs.okstate.edu (STOLFA ROLAND JOSE) Subject: Neat idea Message-ID: <1992Oct26.143400.11545@a.cs.okstate.edu> Organization: Oklahoma State University, Computer Science, Stillwater Date: Mon, 26 Oct 92 14:34:00 GMT Lines: 72 Neter's, I have been struck with the inventive spirit. I have the following proposal for all of you who are interested... We've all heard of the micromouse competition. Most of us have heard of the BEAM robotics competitions (mwtilden, et.al.). Some of us have heard of the "Stiquito" (J. W. Mills). Ok, let's put it all together... I propose to have a teathered stiquito micromouse (bug?) maze competition. Here is how I propose to do it: 1. The playing field be a simple sheet of black flocking paper (used on the inside of some telescopes, available at photo shops & Edmond) as the base, non-IR reflecting surface. The maze path (not walls) are placed on this surface with the aluminium tape that is used for detecting broken glass windows in security installations at spacing's to be determined later. 2. The stiquito would be teathered to a simple controller above it. This would allow quite a bit of computational power, without the need for masive lightness. Ie. you could even use your latest greatest PC (notebook Spark station, yea.... :-) to drive this thing. 3. Some of the sensors that would be allowed would be: A. some simple array of IR detectors/emmiters, slung under the Stiquito. B. Simple grouding strip to the aluminum tape? C. Any other BEAM type ideas (yea, keep it simple :-). Ideas??? General idea is that ALL of the sensors must be carried by the stiquito... Uniqueness points for this project: 1. Fairly low cost. I'll be more than willing to flesh out the rules for all of this, if there is interest. Given the right conditions, this could even be strictly a programming project (provided a PC, a stiquito, and a simple interface). Ie. you could build your own, test your code, and then mail your software entry to some central "competition" site. 2. The first "legged" micromouse competition. The rolling mice have been around for quite some time. Here is a new direction, with it's own NEW challanges, that we can all start on a level playing field. 3. Given that the "interface" to the PC could be activated from any PC language (even BASIC), this whole system might be a good one for getting youngsters involved (even junior high types). Research intrests involved: 1. Yet further research into legged locomotion. 2. Fairly easy system to deal with new control systems software. I guess you could even try out subsumption control systems; Ok, I"ll volunteer to referre this thing. Just send your ideas to "rjs=smm@a.cs.okstate.edu", and if there is any response, I'll do my best to fully flesh out the rules, etc. Oh, and by the way, hats off to Jonathan Mills for the Stiqutio design (really neat) and the BEAM-team (Mark Tilden & ???) for their concepts that glued all of these ideas together for me. Great job everybody!!! Keep those neat ideas comming. Roland Stolfa rjs@a.cs.okstate.edu From: "Jonathan Mills" To: stiquito@xcf.Berkeley.EDU Subject: DRAFT TR363b Part 2 Available Jan 15 Cc: jwmills@cs.indiana.edu Status: R TR363b, Stiquito: Part 2, Sensors and Controllers, will be available from anonymous ftp to cs.indiana.edu *after January 15, 1993* The report describes controllers built during the the fall 1992 VLSI seminar at Indiana University. The report is available in draft until a few late projects are received. The projects listed in the report are accompanied (where appropriate) by additional design files that can be ftp'd and used to burn your own FPGA gait controllers. Or, use the files as "parts bins" to design new controllers. Contents as of January 1, 1993: Stiquito, A Platform for AI.........................M. Scott Transistorized Controllers..........................J. Mills General-Purpose Controller (FPGA)...................S. Pullela Parallel-to-Serial Controller Interface.............Hendry & Heininger Parallel PC Interface...............................Heininger "Real Soon Now" stuff: Interfering Variable-Rate Clock Controller..........C. Reising Two New Stiquito Designs............................J. Mills Sincerely, Jonathan From: "Jonathan Mills" To: stiquito@xcf.Berkeley.EDU Subject: Lukasiewicz Logic in Feb'93 Sci Am Cc: jwmills@cs.indiana.edu Status: R A readable and amusing introduction to Lukasiewicz logic (aka fuzzy logic) and its relation to chaos is given in Ian Stewart's "Mathematical Recreations" column. Substitute position in a Cartesian coordinate system for "truth" and the relationship between Lukasiewicz logic and a gait controller will be apparent. The controller has a family of stable limit cycles which can be selected by forcing a starting position in the leg's trajectory space. Varied gaits emerge by inhibiting and activating combinations of six legs (as on our project with the 2-dof Stiquito leg). Gaits vary within an inhibition-activation combination when noise is introduced, causing the leg trajectory to switch to a new limit cycle. A simulator with a "control panel" and graphic displays is available that runs on a Mac under the WingZ 1.1 spreadsheet. Look for: legs.hqx in the ftp public directory at cs.indiana.edu *after January 15, 1993* Sincerely, Jonathan