Journal97 nov
From Simreal
Contents |
November 1997
Sat, 1 Nov 97
First Trick-Or-Treater came up a bit after 6:00... little kid, so I canít run the monster for her. Next up, maybe 12 year old boy... almost jumped off the deck! I got clumps of kids every, oh, 15 minutes or so. A couple of great reactions! Jump, screech, giggle - the girls had the best response. Guys had to look cool, so they didnít jump much (but I saw them twitch). A couple of groups... after the jumper did his thing, waved their hands around the can, walked up and down the steps, and generally tried to make it trigger again. Sorry, kids, only one show per group. The general reaction (when I got one) was a quick startle when the alarm started, staring to find the noise. But before that settles in, whoosh! The garbage can explodes open with something inside it. Canít really see the monster, but the lid flies up, the can shakes, it's the overall motion that makes the second startle the big one. Yup. Definitely was worried I was going to lose some kids off the deck, but they all made it in one piece. Iíll have to work harder next year ;-}
Ran out of candy about 9:00 (well, I actually ate the last piece myself), so I turned off the porch light and blew out the pumpkin candles. Except for the "Welcome" pumpkin, sitting quietly on its switch. The bait pumpkin. I'm still hoping to catch a smasher.
About midnight, I pad out onto the deck in my robe and cold feet. No smashers. <sigh>. I take in the equipment.
By the morning, the smashers had taken the welcome pumpkin (which is why it doesnít show up in the photos). Lame. I mean, really, totally pathetic. How much nerve does it take to sneak up on a dead house, lights out, Halloween over - and take a pumpkin on the edge of a deck. Heck, they saved me the trouble of carting that particular pumpkin out back to the compost pile. Real pumpkin smashers try to damage the halloween experience itself. They act during the trick-or-treat times. That takes nerve. Guts. I dare you. <heh> <heh>.
Mon, 3 Nov 97
I was expecting my manifold in the mail today. Instead I get a phone call from the supplier saying they put it in the mail today. Iíve taken to the drawing board again, touching up my Boris designs. Looking pretty okay, here. As soon as I find a decent CAD or drawing package (or find a good use for one I have), I will start putting up detailed Boris specifications. "Build-Along With Boris". Should be a fun page!
Wed, 5 Nov 97
Yesterday the manifold came by UPS! Darn, they are quick sometimes. Havenít gotten the call-tag to return the other $3.00 part (just silly... cost more to ship it back than to let me keep it. When Allied got me the wrong part, they just sampled it to me and sent the right one). I also got a couple of 10-32 fittings for the manifold's smaller ports.
So I hook up the trusty-rusty leg model to the new valves. Download some new test software (which, by the way, is easier to run with my Halloween control panel). Poke buttons, prod switches, fire up the compressor (blew the circuit breaker again... have to remember to turn off the heater when I turn on the compressor).
Well, these valves performed excellently. They are fast. Really fast, and I could drive them as quick as I could stand it. No timing results handy, but they really really screamed.
But I could notice the reduction in airflow... the leg seemed a bit more sluggish to respond, and it actually seemed weaker too. For weakness, I already plan to upgrade the thigh cylinder to a 2" diameter version ... and maybe change the leverage a bit in the redesign. The 2" cylinder should give me about 300 lbs push at 100psi air, with the 3:1 reduction in leverage (estimated) I get about 100 lbs push at the toes. Maybe. Marginal, but hopefully sufficient.
I could live with most of the valve's oddities, and I repeat these Clippard valves behaved better than advertised. The problem is in the valve-circuit configuration. The fast valves on the manifold donít exactly replace my 3-position Parker. They have an open center position, where the Parker has a closed center. This functional difference is the killer... when the Clippard aggregate is "off", air is escaping all over the place, drift, consumption, bad stuff. The Parker, in "off", just holds tight.
Iíll snuggle up to my catalogs tonight and see if I can find a niche in this problem to get the perfect solution. I want Clippardís speed with Parkerís closed center. Just got a catalog of Mac valves... hmmm....
I need to decide soon. I am anxious to get back to finishing the mechanical designs. I am playing with a new CAD package that may make it easier for me to create Web-publishable drawings. Or maybe Iíll just model it in my renderer and add dimensions in the paint program. Weíll know soon!
Now, to finish taking that roll of film so I can post the Halloween pictures.
Sun, Nov 9
Itís been a busy week, but I havenít really gotten much done on Boris. Well, I have, but in an indirect manner.
I keep meeting people who are excited by this project, so I think I might be able to dig up a local partner sometime soon.
After thinking about it, I am definitely going to put up technical information on Boris, so keep your eyes peeled for the Build-Along With Boris pages! I tried several CAD packages for this, but they were all either a) expensive, b) terrible, or mostly c) both terrible and expensive. The drawings will be done in a package that I found Friday called Delta Cad. It is easy to use, capable, and very inexpensive. To get the full impact of the BAWB pages, you will need to get a copy.
Iíve spent the last couple of days working with DeltaCAD and transferring drawings into it. By the next post, I should have enough done to make a decent BAWB page, as well as get back to making metal!
My other thinking this weekend has been on how I can finance the next phase. Once I build the final set of 6 legs and the body, I need to drop 18 valves and 18 cylinders into it - at a cost of about $2,000. This is complicated a bit by the minor detail that my jewelry income has dried up for the next month or so, while the shop weathers its slow sales season.
My solution? Grovel and beg! No, um, thatís not right. My solution? Sell stuff! Get yerí Boris Shirts and Hats here! That, and I can make the Build-Along With Boris section a teaser - give the blurry overviews for free, but for the actual data files and measurements, not to mention assembly instructions, you gotta get a "Technical Membership" to the Boris e-List. So for $50 or so, you get updated with the latest Boris CAD drawings and notes until the competition. Any thoughts? Email me!
I have a busy list of "todo" items:
1. Time to mail-order the discretes to build my computers.
2. I am going to do software development for the robot soon, especially as I collect the money to install the pneumatics. For this, since I won't have a full Boris on hand, I am going to build a rolling robot. Yes, wheels. Ugh. But cheap. I can test some basic software algorithms around learning and control on this first. Yes, yes, I'll post details on Boris Web as they come!
3. I finally finished that roll of film! Next post, Halloween pictures.
Wed, 12 Nov 97
I have quite the set of Boris plans drafted up now... both major leg sections and the leg to body "spider," plus various accessories. It is a bit frustrating to be out of the shop right now, but this paperwork will make the final product all that much better, plus it should make for some good web content.
I am anxious to test the pressure transducers on the leg. I got these really nice (and cheap) transducers (Motorola MPX700 series) in one of my early batches of parts, and havenít had a chance to touch them yet. Now that I have a decent (if not terribly exciting) solution to the leg-angle measurement problem, the muscle sensores are the last to be implemented. I was going to do some work tonight, but with my wife sick this week, I am going to put it off in lieu of the more important task of taking care of my sweety.
Thu, 13 Nov 97
A few days ago I ordered the stack of electronics I needed to build the 10 BotBoard2s. Things are cheaper in bulk! Anyway, I got two of the three packages today. Zowie! Jameco and JDR Microdevices are very responsive. For that matter, the third supplier is DigiKey, and they shipped the same day also... they are just farther away. Their online link to the UPS tracker says I should get the package tomorrow (well, my interpretation... it left the last reasonable tracking point today, and should be making a beeline to my city tonight).
Very exciting. I will build just one BotBoard2 computer first, to debug the process. I will then use this in the mobile robot on wheels that is my software testing platform (and Christmas or Birthday present for my kid). For that, I got a dual-motor gearbox device from MondoTronics that should get here eventually. I just love mail-order. Now, of course, I am broke again.
Tried the MPX700 pressure sensor tonight. Works! I always feel like a proper mad scientist when I am doing the electronics experiments, I donít know why. Anyway, this has got to be the worldís easiest sensor next to a potentiometer. Stuff 5vdc into one pair of leads, and read a 60mv or so output range on the other pair. Tack on a capacitor, and voila! Clean linear pressure sensing. I watched this on the scope for the trials. Now I need to make an amplifier to take the signal from the 60mv average range with DC offset and turn it into a readable signal from 0 to 5v. A little DC offset, a little amplification. No sweat. I hope. Now where did I put those reference books?
Another nice thing about the MPX700 is that it is rated for linear output from 0 to 100psi, but the overpressure rating is up to 400psi. Wonít burst until 700psi or so. Should be plenty robust.
Actually, my only concern is keeping the air line from popping off the sensor.
Sat, 15 Nov 97
Found some bugs in my blueprints. Fixed them. Still trying to design a good body for Boris. Check out the Pictures pages for the first drawings!
This weekend I am working on the BotBoard2, and next week I will be working on the new boris legs and/or the initial software technology to run it all! It may seem early, but time is trickling through my fingers here. I am going to start software in parallel with hardware, to try and keep up. Especially since the financing for the main pneumatics may take a while to put together -- I have to keep moving on one front or another if I hope to make it in time.
I haven't made my preferred Friday update the last few times... I'll see if I can get back on schedule. One of the big hassles is in photograph management. Photos are spendy! Plus, the time to go out and get them developed, scanned, processed, blah blah... anyone want to donate a digital camera? Anyway, in spite of this, I finally added the Halloween pictures to the picture pages!
Mon, 17 Nov 97
Yesterday, I had a wonderful time soldering together the first BotBoard2. It went together smoothly and easily. Didnít dare try it though...
Tonight I plugged it in and turned it on... no smoke, a good sign. Then I fired up the pcBug11 package, and it seemed to communicate with it. I guess it works!
Well, hard to say. Doesnít do anything yet because I still have to learn to program it.
The pcBug11 is an, ah, utilitarian package. I am researching other languages, packages, and systems to support my programming. Iíll have a complete list here with links, eventually!
For this toy robot testbed, I need to build a few features. I have two motors to control on the output. I can add a couple of whisker touch sensors, say front left, front right, and backup. I want to add a speaker and one or more light sensors.
I donít know how I want to do the brain yet. This weekend I will dig out my notes and compile an overview for the testbed brain. I want to give it a bit of personality. It will need to have some form of internal goal that keeps it moving about, and some form of problem solving. I want it to be a learning system, with decent memories and the ability to change behaviors based on feedback. This, of course, is why it needs internal goals - it canít tell what is working or not in its behavior unless it has goals, likes, and dislikes. These give the feedback to modify the systems.
I want to explore several technologies: Fuzzy, Genetic, and Adaptrodes.
Tue, 18 Nov 97
There are now several active tracks of inquiry on Boris: The body design, Testbed toy software, and soon Testbed toy hardware. Plus, figuring out how to work the HC11 MCU for it all!
I spent a bunch of time at the PARTS and Motorola sites finding out more about the HC11. I found a Tiny Forth, some C languages, some assemblers, some communications tools, this and that. The pcBug11 tool seems to be the main utility to program the board, but it is very primitive. I tracked down the documentation for the bootloader on the HC11, and will probably build my own interface. Now, I am wrestling with the relative merits of Tiny4th and assembly language.
I also worked out some more details on the Boris body design. Having trouble working out the space-frame where the cylinders are mounted in the center, and the heavy parts mounted above. I may end up making two bodies, the first as a prototype, and the second to iron out the bugs. Iíll at least have some body pictures up on the BAWB pages this weekend.
Thu, 20 Nov 97
Had a moment of great doubt last night... measured the down-force on the prototype #2 leg. Not as much strength as I expected (with a 1:6 leverage loss, instead of the 1:3 or so that I expected). Must be some losses that I am not anticipating. Could be the scale I am using (cheap bathroom scale).
I also weighed the leg assembly and cylinders - under 20lbs! Good .
Okay, so worst case I assume I have a 1:6 loss of strength due to the design (though, to be honest, design #3 is different from prototype #2, and I should get better efficiencies).
On a 300 pound robot, there must be 300 pounds of up force to keep it off the ground. That means 50 pounds per leg if all six are on the ground, 100 pounds per leg using the fastest tripod gaits. But not all 300 pounds are supported by the cylinders - some of that is resting directly on the ground through the feet.
Now, with the theoretical 314 pounds push on my 2" cylinder, I should have about 100 pounds down-force at the theoretical 1:3 leverage loss. But at 1:6 I get closer to 50. That would mean, as soon as I pick up three legs to walk, the body starts to sink.
Okay - there are three basic cases to worry about.
1. I have enough strength to do the tripod gait. No problems. Kicks butt.
2. I donít have enough strength to do tripod. So, the robot bounces as it walks, or it uses a slower gait that leaves more legs on the ground. Still not bad -- just slower.
3. It is grossly underpowered, and can barely hold the body off the ground. That would be sad. So I would drop casters under the body, and drag it. One way or another, I got to Robot Wars '98.
Maybe I can upsize the thigh cylinders again. 2.5" gives me 491 pounds of thrust, at 1:6 worst case about 80 pounds down-thrust. Hmmm. It would be slower and more expensive. I think I'll stick with the 2" for now.
Now, there are two other axis of motion. The hip swivel and the knee bend. The knee will only give me 30 pounds or so of force, but if I keep this leg section vertical to the ground I donít need very much (it gets carried by the joint instead). Just enough to push/pull the robot around. Same on the hip.
Even if I am forced to drag the body around, those legs are wonderfully fast, and the force of impact is extreme. I can drop sharpened points on the feet, and do some serious damage to the competition (well, at least Iíll scratch their paint).
As I am doing the motion drafting, I found I made the body and spider too small vertically. Back to the drafting board! I hope I get it done in time for this update...
Fri, 21 Nov 97
Well, I had to work hard, but I got all the pictures up into BAWB I wanted. I now have detailed drawings for both leg sections, the re-worked spider, and the body. Plus basic power-plant layout. Most of those details are kinda hard to see in the jpg in BAWB, so you'll just need to buy the subscription I guess.
Talked to someone out there in Robot Land today -- might be able to get me a corporate sponsor! That could make life easier.. but I'll believe it when I hold the money in my hands
This weekend, I can take a break from the main Boris drafting and work on the Testbed toy. I think I will split my time between software research and hardware setup. Got a few chips to test to drive the motors -- think I found me a simple, cheap solution to the H-Bridge problem for small toy motors (more later! Just think analog switch...).
Every day or two I get another e-mail from encouraging and/or helpful readers. Thanks guys! (and, well, that isn't a sexist statement, 'cause no gals have written yet...) Keep up the e-mail, it really helps keep morale up. And also, those occasional e-mails pointing out inefficiencies are handy too.
Wed, 26 Nov 97
Whew! Iíve been running in circles on the software front - reading about neural and fuzzy systems, sitting down and inventing architecture for Boris, getting stuck. Stepping back, and thinking about the Toybot. Finding new insights. Going back to literature, reading more information, trying to mentally apply it to my architectures and ideas, round and round we go!
My brain hurts.
I have 10 pages of notes now. I have another stack of papers to read, and Iíll probably add yet another stack of pages to my notes! The fun part is going to be condensing the notes, and making something legible for you-all to read . Of course, I donít expect my initial notes to describe anything wonderful right off the bat... but they will describe a direction. I will probably need to write a bunch of software to test and simulate everything...
... speaking of which, I wrote a fun little single-neuron simulator once, a long time ago. It did fair job of emulating the Hodgkin-Huxley equations, but that seems to be long lost and gone. Anyway, most of those neuron simulations are heavy on the math - mine used only addition, subtraction, and division by powers of two (shift, for the bit-heads out there). Not very fancy, doesnít learn (but could probably learn by adjusting the threshhold values), but does a fair emulation of a neuron. Out of nostalgia, I built another one. Not as elegant, but it works okay. Check the Build-Along with Boris area for more information.
Thu, 27 Nov 97
Turkey Day! Went to my dadís, ate bird until I passed out. Had fun!
Sun, 30 Nov 97
In spite of the holidays, I got a fair amount of work done. A bunch of it doesn't show here yet, but it will trickle in over time. Check out the BAWB section for my current software technology notes, and some toy sims.
(Editor's note: the toy neural sims are no longer posted)
