Journal99 mar
From Simreal
Contents |
March 1999
Thu, 4 Mar 99
Within hours of last week's update, Rick's parts came by big brown truck... and they reflect his usual high quality of work. I then spent the bulk of the next week putting together the hip-to-body joints (see the pictures page).
To get decent alignment on the four flanges that make up a joint, I weld them all onto their frameworks with a snug-fitting rod passing through the system. Note, I also must include a bushing in the relevant fittings, to get that snug fit.
The heat of welding slowly destroys the bushing, which is okay since they are cheap and I have spares, but more important, it builds up an oxide layer on any metal that gets too hot.
In my first hip, I used a single rod through all joints, and my alignment was a little bit off so that rod was fairly snug. After everything had cooled, the rod was so stuck I had to cut it apart and pound it out of the fittings with a hammer. The joint works great, but the pain and suffering factor was a bit high.
On the next joint (and all others thereafter) I made two changes. The first is, I used a two-part rod with a well-fitting tube to lock the halves together. Each rod goes through a flange pair, one at each end... and I can then see the mis-alignment of the system by which way the ends point where the come together! After some delicate pounding with my trusty hammer, the ends get fairly close and I slide the tube across the gap to make, effectively, a solid rod. This works very nicely.
The second change was to add some durable lubricant to everything, to discourage the buildup of oxidation (maybe, I don't know) and to help prevent the metal parts from seizing together. This all seemed to work fine, and the rest of the hips went together very nicely.
On the software front, I did some extensive work on the in-circuit programmer. This was far more of a struggle than it should have been, but now I can program both the Flash memory and SRAM, with full verification and whatnot. I am just now wrapping up my AVR floating-point math routines (ugh), and I should have Fuzbol showing signs of life very soon now.
March 13 is screaming up at me, and I still want to have something on display at the show even if it's not Boris proper. Plan-B is to hook up the proto-3 leg to one of the new hips and add some manual controls to the pneumatics so an operator can make it wave around. Not computer controlled, but life goes on; *I* know I can do it, because I have already! The problem is that the software I used for proto-3 is now the property of the people I wrote it for...
Tue, 30 Mar 99
The show on the 13th went well -- I made some contacts, though since they haven't gotten ahold of me, maybe I only *thought* I made some contacts... the old Boris leg ran nicely under manual control for the entire four days. I made a few mounting changes that eliminated some annoying interference points and also validated some hip design changes I had in the works.
Rick as sent me more parts, and they are beautiful. Just a few more flanges in the mail, and all of the legs get welded together. He also worked up a new design for some of the cylinder mounting brackets -- a wonderful and simple clamp-on arrangement. I hope to get those into CAD eventually...
I've done nothing but Fuzbol since the last update (with the exception of the show and making up the sensor mounting brackets) -- and I mean it -- pretty much a solid two weeks of 12-hour days, with minimal distractions from work debugging. My day job's release has been a little rough, but with very few bugs that I had to fix. I'm so burnt out now, but the entire Fuzbol language is both operational and has gone through initial debugging. I only have to add interrupt management, and it's fully functional.
I'm going on vacation around Easter for a week, back in Oregon. I'll be bringing a laptop, so I hope to write up the documentation for Fuzbol on the trip -- when I get back, I hope to make the next update a software extravaganza, with the full Fuzbol language and doco. Maybe even some schematics for simple Atmel controller projects, and sample code. Or maybe that all has to wait for May.
All of this is going into my book -- plus more. I may even get that finished by the June 1 deadline, if I work hard. Fuzbol was the hard part; I spent a huge amount of time there. Even so, it isn't fully polished -- the error handling sucks, and many aspects of the compiler are fairly primitive.
I see the light at the end of the tunnel finally -- the mechanicals are coming together beautifully, the software foundations have been laid, most of the expensive stuff has been purchased. The last big problem to solve is the central controller hardware, and the actual software to drive the robot. But even those shouldn't be too awful compared to the many other gauntlets I've run during this project.
