Archive for January, 2008


The ÃœB4RSCOOTER returns to MIT

Jan 29, 2008 in MIT & Boston, Stuff

For a little while. Over the course of the first semester I was involved in the construction of a prototype folding motor scooter. My contribution ended up being part of the control system for a demonstration stand featuring the primary innovative technology, an in-wheel motor-suspension robot doohickey. The majority of the machine was built in Taiwan, and was shipped from there to the Milan Motorcycle Show in Italy. Only now has it returned to the Media Lab.

Plans to just set the crate on fire were rejected.

 The initial decrate-ifying, which was mostly just ripping off one side of the crate using whatever brute force was available. Normally the panel woudl come off in one piece, but the wood was so thin that it just disintegrated.

There was no fog effect or bright lights. Nor trumpets. Nor suprise confetti.

Ooh, shiny. There it is.

You should have seen the ones at the show.

 As with any proper photography of machinery, there must be women. Conveniently, some from the scooter group were present.

Size doesn't mat....wait a sec.

Size comparison to another motor scooter that was conveniently parked outside. It has the same wheelbase, but appears smaller… mostly because it’s SOLID FREAKIN’ METAL. Instead of that cheesy bubble-shaped sheet metal cover over some tubing. Despite the heavy metal, the whole machine only weighs in at about 120 pounds.

So what’s going to happen to it while it’s in the lab? Control system development. Wheel robot development. Battery experiments. Random upgrades. Hoonage. Stunt jumping, etc…  And I will of course stay with Smart Cities for this semester. Whee!

TB4.5MCESP1 Update 12

Jan 28, 2008 in Bots, Project Build Reports, Test Bot 4.5 SP1

Yes, another one! I wonder how many build report posts I can rack up before the bot’s finally done.

Completed arm motor assembly. This is a stock drill motor with the casing removed and with a modified output shaft. The motor is a 540 size – slightly shorter than normal drill motors. With the intermittent duty cycle of the arm, I decided a shorter, overvolted motor would perform just as well. I had to file down the endcap of the 550 drill motor in MCE’s arm gearbox such that it would clear the drive belts. With this, I won’t have to do it.

Test mounting shows that most everything is where it needs to be. Unfortunately, while deepening the right arm rail hinge pin hole to allow the final assembly to sit at proper width, the UHMW managed to flex out of the clamp and get sucked into the drill bit, causing one hole to be drilled 5/16″ all the way through. This doesn’t impact the arm operation, but does lose me a fastener location.

UHMW has a propensity to suck itself into whatever cutting tool you’re using at the time unless it’s REALLY well secured.

Hinges drilled. You may notice that they are missing the pins. I intend to use some hardened 3/16″ steel rod as “pins” instead of the cheesy stuff that came with the hinge, which I could bend a 4 inch section thereof by hand. The hinge seams will also be tack-welded shut to use the extra strength. Also, I might incorporate “nutstrip” along the inside of the bot at the attachment points to lessen the likelihood of losing a wedge even more.

The almost-done frame for the überswitch. In MCE, this was a hack with some aluminum angle, washers, and spare standoffs. I’m giving the switch a proper frame in 1″ aluminum channel milled to the appropriate shape. The key slot hasn’t been made yet, because…

…I couldn’t  find a 1/8″ endmill.  There was a broken one, but that doesn’t count. I did find this tiny 2mm endmill that I briefly started using to cut the key slot before realizing that the machine can’t run anywhere close to the speed the cutter needs. Something like this needs 5 or 6000 RPM or more to properly use. I stopped before causing some damage or breaking the tool. MITERS has 1/8″ endmills for the Bridgeport, so I’ll swing by some time later.

A few days ago, I decided (after almost blowing up my 12v power supply from a dead short) that I really needed to integrate my new charger and its power supply into one unit, along with accessories, to prevent from having a clusterfuck of wires every time I need to charge a battery. The prototype was made from GIANT LAZER’D acrylic, but acrylic does not cold-bend and I couldn’t find a strip heater. So I used the prototype as a template to transfer the hole pattern onto some .1″ aluminum.

The sheet was then folded on the Media Lab sheet metal device. It’s only built for 20 gauge steel, and here I am trying to bend .1″ thick aluminum on it. It did NOT go smoothly. But in the end, what brute force I could dish out did work, and there were no heartstopping cracks or odd noises from the machine.

Test assembly. I’ll probably mount terminal strips, fans (the charger does get hot despite what its 40mm fan tries to do) and other accessories as necessary.

Great, now that almost everything else is done, I just NEED SOME TIME ON THAT WATERJET! I don’t have a ML network login and so can’t hop on and do it just any time. Fortunately, the nice folks as the ML are always willing to help out. We’ll see where this week leads.

Bot on!

TB4.5MCESP1 Update 11

Jan 27, 2008 in Bots, Project Build Reports, Test Bot 4.5 SP1

Two-day update! Building over the past week must have completely worn me out, because last night/this morning/today I managed to sleep for 14 hours straight. In fact, I flopped down and fell asleep right before starting on the last update. So here is a combined two-day update.

It’s coming together. In fact, after I get some intimate time with the waterjet cutter, I pretty much have to just drop in the parts (and file, tweak, hammer, etc….) and wire up. Everything’s still a jumble of parts, but there should be a fair margin of time before Moto left for practice driving and tuning.

Just a random thought.

If you ever wanted to totally rip me off, here’s your chance. This is the 2D flat pattern for SP1′s arm parts that will be handed to an abrasive waterjet some time soon, and hopefully by the end I’ll have little bits and pieces of sculpted aluminum. Note that the arm links and linklets are connected by breakout tabs. Parts that small stand a chance of falling into the waterjet tank, where they will be lost forever until the next cleaning, so connecting them with thin tabs to other parts or the main sheet of material prevents this.

I wonder how many would-be parts they mine out of the tank every time they clean it out.

Oh, hey, let’s make a cardboard robot.

I got some time on the Media Lab’s GIANT LAZER and made some prototype / sanity check parts out of cardboard. Not too exciting, but I’ve seen people cut things as odd as graham crackers and chocolate bars on it also (the ML smelled dessertful that day). Especially important is the EBay pieces, since it lets me know that everything does indeed fit IRL.

Here’s a video of it in action (Metube)

Putting the test assembly through a somewhat-assembled frame.

Another precision product of the MITERS lathe, the inter-arm-beam-standoffs! I have no clue what on earth happened here. The drill bit sure looked like it went in straight and didn’t wobble on start. I suppose not.

Just to be extra-careful on the remakes, I didn’t drill the whole length of the part since it was going to be a standoff anyway. I also tried using a center drill to start the hole. Go figure, there was no center drill available, so I made do with a countersink. That somehow worked well.

Here are the arm standoffs along with new hinge pins. Only after I disassemble the arm did I discover that in fact BOTH bottom hinge pins were bent, not just the front as I thought. A good punt directly to the arm from a drumbot will do that, I suppose. I might try casehardening the pins using the oil-dunk method, but suspect they might warp. If the MITERS lathe could machine the steel using chipped and dulled tooling, it probably isn’t very hard.

Hinges cut out and milled to (somewhat) the right size. I was off by one digit in the McMaster part number and got a hinge that’s slightly wider than what I need. Not wide enough to warrant a return , but wide enough to interfere with the design.

So the solution was to mill them down to size. In another episode of Charles fucking hates flakey public tools, the ML mill has a severe issue with Z-axis drift and backlash. I swear I cranked the dial to exactly what I needed every time, but the worst inaccuracy is still .05” short of what I actually needed. Fortunately this .05 does not interfere with the design, and I didn’t feel like bandsawing off another section of hinge by that point. Half of each hinge is at the correct dimensions +/- .002, and so it’s acceptable. I wonder why those parts were consistent.

Maybe if I had checked my progress as I lowered the Z-axis for another pass I would have saved alot of headache as I would have known that I was about to fly way past the stated dimension. Oh well.

I was tired at that point and decided to save the drilling for another day.

…but neither tired enough nor pissed enough at the mill. Finding a 1.5″ square rod of 2024 aluminum just gives one more botting energy anyways. That’s exactly what happened – while walking out of the lab, I discovered the 3 foot chunk of Al in the material rack and immediately decided to extend my hours.

Here’s the arm gearbox mount milled to spec. The 1.5″ thick bar took almost 15 minutes to cut on the horizontal bandsaw, mostly because it’s also old and creaky and I had to set it to go REALLY slow or the blade would jump the pulleys, resulting in bad.

For these cuts, I set the Z-axis lock a bit such that I could feel the backlash in the handwheel before it moved the axis. This was a godsend.

Drilled, threaded, and countersunk where appropriate. The finished arm gearbox mounting plate, made of a chunk of 2024. I wonder what else on the bot is 1.5″ square, becuase I want 2024 mounts across the board now.

The 5/8″ hole was made using my ghettoedging method, even while the edge finder was staring me down in the endmill rack. It worked well. An added twist this time was to run the spindle backwards such that the tool contacts but doesn’t cut.

Oh, hey, maybe I can remake that dumbass drive gearbox middle plate.

Front end of arm gearbox assembled. With retaining ring and everything. There are a total of THREE bearings that will be supporting the extend-o-shaft this time. That means it will probably bend and ruin my day anyway.

So that concludes the work as of tonight. I might try to fit the chassis together soon and maybe wire up for the odd test drive or two.

Oh, and I also found a 1/8″ plate of 2024 in the scrap bin on my way out after milling the motor mount. Fortunately, I could think of no parts that involved 1/8″ plate, and so did not stay a second time…

Maybe tomorrow.

TB4.5MCESP1 Update 10

Jan 25, 2008 in Bots, Project Build Reports, Test Bot 4.5 SP1

I will never take apart TB’s gearboxes again.

I will never take apart TB’s gearboxes again.

I will never take apart TB’s gearboxes again.

I will never take apart TB’s gearboxes again.

Did I mention that I will never take apart TB’s gearboxes again?

Today was a combined 8 hour work session, three of which were spent trying to put the gearboxes back together.

I took the GBs apart to swap out the almost 4 year old motors. I got these at Dragon*Con 2004, and they have seen service in 2 builds of TB and also Trial Bot 1.0′s weapon assembly. The brushes are shot to hell and the armatures are half toast. In the same session, I also bored out and installed a new 2″ Colson wheel to replace one of the rear drive wheels, which has a large chunk missing. The rest of the wheels are still smooth and round, so I did not replace them.

The new motors are real Johnson power tool motors (Serious Businessâ„¢) and are rated at 12 volts. Since the bot’s running 15v this time, I decided the old 7.2v motors were probably going to go poof with a minute’s driving while hurling the bot along at Warp 2√π.

So the problem was putting the gearboxen back together such that they still worked. These gearboxes have three main parts – the motor mount plate, the intermediate stage bearing plate, and the outer bearing/mounting plate. Only one was made by a repeatable process – the outer plate, which was CNC’d. The rest have slight variations in hole placement and dimensions.

Suddenly, holding a jumbled pile of gearbox parts, I forgot which part went to which gearbox.

In two combinations of three items with one held constant in each, there are four possible arrangements of parts. To add even more complication, though, the middle plate is square and symmetric (supposedly) with four possible orientations. Had the parts been milled or made with more exacting specifications, this won’t matter. But unfortunately there is only one or two positions of the middle plate in which any combination of the above parts would work (otherwise the gears run into eachother and bind). So I was now dealing with up to 16 ways to put the gearboxes back together.

Great. Whatever. In the end, one works fine and the other still binds some. The interference appears to be from the first stage, and rather minor, so I’m tempted to just let it wear itself in.

Odd wear pattern from when I took apart the left drive motor. The HF motors have 3 planet gears. This is a 5-pointed wear pattern. 5 is not a multiple of 3. Can someone tell me what crazy harmonics are at work here? Is Satan in possession of the bot? Good.

Milling around. This is the center “cutout” where the arm sits. I couldn’t find an edge finder, so I did it up the ghetto way – coat the end of the piece in permanent marker, then crank the axes until the edge of the cutter just barely scrapes the coating off without actually cutting into the piece. Conveniently enough, your cutter is already in the spindle, and you can zero the dial / DRO and go. I got amazingly close by this method and handwheel dial reading alone – 3.247 on a 3.25″ long cutout. That’s better than the UHMW’s manufacturing tolerance, so I’m happy.

I plan to use this ghetto-edging tactic more often. Sure beats digging out an edge finder, making sure it runs dead concentric, then accounting for the tool width offset, then taking it out and putting in a cutting tool…. maybe on a machine without a mile of backlash.

The bot as of now. It’s looking like something. Holy snorf.

On the way is a 30 pound package containing one (1) 12″ x 39″ x 1/4″ 2024-T3 aluminum plate and one (1) 12″ x 28″ x 1/2″ 2024-T3 aluminum plate with one side mirror finish (ooh, shiny). That’s enough aluminum to last me a long time. I got all of this for a bit under $100, or about $2.8 per pound of good aluminum. That’s a killer deal, and some of that UHMW might be replaced by some heavy metal soon.

It’s coming together! Bot onwards!

TB4.5MCESP1 Update 9.

Jan 24, 2008 in Bots, Project Build Reports, Test Bot 4.5 SP1

After a marathon 12 hour work session, SP1 is now approaching the “looks like something… kinda” stage in construction. The frame holes are now mostly drilled. The spindle for the arm gear is also complete. Some big aluminum plates are now on their way (the front desk is going to hate me in a few days) and flat patterns have been made for the waterjet and GIANT LAZER.

I still haven’t figured out how I’m going to get to the event. Go figure.

YES! I'm not in California, it doesn't cause cancer!

But before we begin with the usual monotonous build pics, here’s a picture of a lathe headstock bearing.
After one of my “.375 inch” drilled holes came out .393 on one end and .420 on the other, it was obvious something else was up. If you’re keeping track, that’s a 1.3 degree taper across a hole about 1.125″ long.

The MITERS lathe has a severe case of tailstock misalignment. I turned two pointy centers, put them end to end, and discovered that it was off both vertically and horizontally by about 1/16″. That’s insane. Naturally, I must think of ways to remedy this that involve random hacks and waste my time.

  1. Crank the tailstock adjustment screws, the first obvious choice. They haven’t moved in 40… 50 years? Since 1500B.C.? It took Herculean effort, but I was able to eliminate the sideways misalignment.
  2. Read the instructions on how to lower the spindle by dropping shims from the headstock bearing
  3. Attempt to, but pop out a bearing spacer (The bronze component hanging awkwardly from one end)
  4. Try to pop it back in. It flies out and lands somewhere in the inner cavity.
  5. Take half of headstock apart and dig around interior of machine. Fumble through mountain of chips, drillings, and curls.
  6. Find it, jam it back in, and pretend that never happened
  7. Fold pieces of copper foil to the correct height and wedge under tailstock ways. Works momentarily before foil compresses and alignment is off again
  8. Jam bigger piece of foil under tailstock ways. I’m stopping, I don’t care any more.

And that’s how I blew two hours and ended up covered in 60+ year old machine grease. That must be really good for me.

Anyways, buildpics.

When you have no cutoff tool and need to get stock of a relatively small diameter cut to a certain length, what do you do? One way is to hacksaw it, but I couldn’t get the hacksaw to stick to one spot on a threaded bolt (the only source of 1/2″ non-hardened steel I had). The “hack” was to take a biiiig drill bit and just drill the thing down. This cut was consuming WD-40 and turning it into thick white smoke as fast as I could spray it on. In the end, it worked.

The finished product of the machining session – A hollowed out drill spindle and the EXTEND-O-SHAFT 2000â„¢. I moved the arm drive gear farther out to the middle of the bot this time around, and so the normal drill spindle as-extracted from a drill won’t reach.

By the way, I turned a 6″ long 1/2″ diameter steel bolt into that stupid 1.5″ rod the gear is mounted to. A bit of a waste, but hey.

The assembly.

In another episode of “Charles rigs random hacks to fix mildly abused public tooling”, here’s a quick rigged fix to the top end of the Media Lab mill. The mill head can be adjusted by a crank which drives a rack and pinion. The problem is that the head liked to get stuck in an elevated position and the black metal cap, which normally stop the rack and lets it exert leverage to get the head back down, broke off. And thus the mill can only get higher.

Someone had tried to drill a hole and using a set screw-like setup to hold the cap down, but it appears to have failed and cracked the cap. The solution was to drill through the big cast iron column and drive the screw through it. Now the rack butts against a 5/16″ cap screw, which allows the head to move back down again.

I didn’t even get to the milling part this time. Oh well.

I hate retaining rings. Absolutely hate them. They’re a pain in the ass, need a special and not-common tool to work, and like to fly off into random directions and get lost if you don’t work said tool with utmost precision. I would never, ever use retaining rings on anything I build.

But Charles, isn’t that a retaining ring on the drill shaft? Yes, it is. I used a retaining ring to hold in the bearings on the Extend-O-Shaftâ„¢ and drill spindle. In the end, it’s still the most compact way to keep something axially aligned on a shaft. This was yanked from the pile of drill parts I always keep around.

Frame holes mostly drilled. This whole process tool about half an hour, since I had already center punched the holes. UHMW is very soft and forgiving, so the drill process was as simple as setting the depth stop on the drill press and slamming it down on each center after a very quick pause so the bit can align itself with the center punch dimple.

The dead axle pin holes were made flush on the bottom with endmills in the drill press spindle. Countersinking was also smooth and painless because of the depth stop.

There’s one fuckup on the front left side where I let a piece escape and it got sucked into the 3/8″ drill bit I was using to start that axle hole. So now there’s a 3/8″ through-hole where there isn’t supposed to be. Fortunately, this part is identical to the one I made for MCE. I might just swap it out.

Okay, so now that the tools are less dysfunctional, I should be able to build quicker. Stay tuned for more updates, and bot on.