Journal98 sep
From Simreal
September 1998
Wed, 2 Sep 98
I didn't get nearly as much done this last week as I had hoped. On Friday, my day-job finally cut the CD for our set of products, and all of the last-minute stuff kept me hopping. Then, my other contract in the evenings has been just dragging on one bloody footstep at a time towards a vital demo (which has been postponed two or three times now)... and I just haven't had the spirit to make much Boris progress.
I did cut 42 pieces, though. I would have cut more, but my abrasive disk had worn down so small that it couldn't reach far enough back to cut the 45-degree angled cuts. Went out and bought a new disk, and had my Oxy/Acetylene tanks filled too. Hmm... 42 of 168 pieces is exactly 25%. I only spent an hour actually cutting (maybe more), with the rest cleaning up the shop, staining the metal, etc.... in a good weekend I can finish the other 75% and start working on either jigs or just directly welding the framework.
After talking with Davis Tech's guy who is working on the Boris parts there (and after getting some side-estimates from Charles Pitzer's machine shop), it seems there isn't much chance of my getting all of the Boris machined parts done in September. Once I get the leg frames welded up, I'll need to start at the bottom of my list and start hand-making parts... hopefully to meet the shop in the middle somewhere in October. Not only are there the 168 pieces to the Boris frame, but there are 168 brackets and flanges to either machine or fabricate. If anyone wants to donate $4,000 of machine-shop time and materials, not to mention working 200 extra hours on top of your day job in September.... call me! I've also got this bridge in Brooklyn that I could trade for parts.... ;-}
The word "ambitious" keeps coming up with regards to Boris... seems to be a trend.
This week sees me in a design phase for my day job... and it should also see the completion of the demo for the other work. I may have a free weekend to make some serious Boris progress! Not to mention, some energy in the evenings to do the electronics design work I've been putting off.
I still want to get Boris fully assembled before Halloween... but that seems to be a receding goal, with the slow turnaround on the machined parts. Regardless, Boris must be running by the new year.
This month marks the 1-year anniversary of the Boris Project Web Site. Progress was rapid at first, with prototype after prototype being turned out, and seemingly rapid advances on every front. After one year, and a cross-country move, I feel downright sluggish. Now that most of the details have been worked out, I am moving into the final construction for the mechanicals, I've purchased the pneumatic actuators, and I'm making a (hopefully) final electronics design. The end is near. Soon, we will all know how well Boris works.
Thr, 10 Sep 98
Finished cutting and de-burring the 168 leg framework pieces... <whew>. Now, to clean off the marker die and weld them together. I'll spend some time designing jigs, though -- if I have to hand adjust and clamp each piece, it will take forever to weld!
I've spent an amazing stack of time fiddling with MCU circuit design the last week -- I've got an architecture around the Atmel 8515 that should totally kick butt, except my version of Eagle isn't quite up to the task of the layout (it limits me to a 3" by 4" board... too small). So I've paid for an upgrade that will double my board space, and launch me into the final electronics.
The brains will be modular... on one board, the Atmel, 32K of RAM, battery backup for the RAM, address demultiplexing, and a dual UART for communications and port expansion. Off of this board I can piggyback another one that will do RS232 level adjustments on the UARTs, give me 16 bits of really fast A/D, and provide space for my MOSFET valve drivers. It is structured around memory-mapped I/O instead of the usual SPI, so there are a bunch of pins going between boards (with locking connectors, of course; vibration will be intense). In theory, given the bus expansion structure, I could chain 64 addressable modules off of this system... but one or two should be sufficient. Oh, it is also setup for full in-system programmability.
I hope it works. I've kinda overextended my skills here... but $100 or so to Alberta Printed Circuits, I'll know in a week or two.
On the "I'm working too hard and can't get up" front, I've resigned from my robotic contract... on friendly terms, as best I can tell. As a side effect, though, I want to re-design my control software (white-sheet it) so I don't infringe on their rights. That will actually help my cause, though, since I've been reading some heavy books on Reinforcement Learning and Fuzzy Programming -- my new design should be quite interesting. In fact, I'm looking to write a new chapter on my "Brains" whitepaper in the BAWB section soon. Should be fun!
Seven weeks to Halloween. Pneumatics should come in a week or two (five weeks pad), electronics and programming will probably take three or four (three weeks pad), and mechanical assembly (eg: welding) may take every second... argh! Gotta check with my machinist and get an ETA... also need to decide if/what Chuck Pitzer can do for me (and how much I want to spend there, vs. killing myself in the shop doing overtime myself). I still feel like a one-legged man in a butt-kicking contest.
Sat, 19 Sep 98
It has been a long time since I've had a Saturday update. Last week the time just slipped by, one day after another, until WHAM! Saturday. Part of the reason I delayed was that I expected one or more good-news items to be ready to go into the journal...
The pneumatic components for Boris are ready for me to pick up now...except for the valve manifolds. So almost good news there. The Circuit boards are designed, but I haven't laid them out yet... maybe tonight, so I can send the files to Alberta Printed Circuits and get my first test board built and tested. I haven't used Alberta before, but they allow me to send Gerber files while the PC Express requires I use their software... and with the Canadian exchange rate what it is now, it's actually quite the bargain.
I bought an air compressor (portable, gas powered, looks nice) but it won't be delivered until next week. I spent time welding last weekend in the shop, but only did preparation stuff -- a jig-stick to simplify my layup, test welding to get back in practice.
I spent a bunch of time playing with some interesting concepts modifying Reinforcement Learning, but so far my results have been inconclusive... some bugs in my software are getting in the way so I need to fiddle with that some more.
So, a week of progress, but no conclusions!
I am spending almost all of today in the welding shop, and I hope to get a stack of work done. I just went back to the 7 April '98 journal and reviewed my "tips" for successful leg welding, and a good thing too -- I was about to do it the hard way again! And on six legs this time, instead of one, assembly-line style. <whew> It's good to keep shop notes, and review them from time to time.
No word from my machinist yet -- but then I didn't call him friday like I had planned, either. Charles Pitzer is anxious to get some business from me... but I'm feeling a bit of a money pinch from the pneumatic purchases right now.
And the electronics is going to cost me another stack of dough soon... you know, I'm almost tempted to buy one of the systems from Tern] instead of finishing my own MCU design; it would probably be cheaper, and certainly a heck of a lot simpler. Hmmm.... but then, I don't get the custom I/O capabilities, which I would have to build anyway. And I've been masochistic so far, why stop now!
Next week (Sunday night to Friday night) I am in Georgia, so no more updates until October. I'll be spending some quiet hotel time while I am gone working on software solutions for the Robot -- even if I don't get Reinforcement Learning concepts incorporated, I need a powerful and flexible robot control language that can handle the I/O, communication, and so forth. I've written one already, but I want to go next-generation now. If I do it right, it will be both simple and it will knock your socks off!
Getting Boris operational by Halloween may be bottlenecked by two things -- electronic control, and machined fittings. Fortunately, I actually have a back-stock of some of the fittings I need in rough-cut form right now... I was hoping to get good ones, but I can probably get most of Boris together mechanically the way I did originally... if I can just get a couple of things out of my shop in time, which seems likely still.
See you in October!
