Robot Army

TiggerBot II is a loser. I probably won't be doing any more work on it.

Evidently the attempt to make an easy-to-build robot was pointless; robots are hard and people are stupid. This is the new plan, tentatively named 'devilkitten'. It will feature many components copied from tb2. Surface mount components and custom machined mechanics are now back in play, however.

I'm having this cut today.

It is sort of hard for canal to give me something out of spec since they don't really have specs to begin with. In any case, when they say 0.25" acrylic, they mean "somewhere in the 0.20-0.25 range". I usually just measure the last thing I had made with the micrometers, assume it's the same, and cross my fingers. Maybe if I stuck to one color it would help. In any case it fit after a wee bit of dremeling. This robot is my second "super traditional" robot, i.e. round with differential drive and two casters, the first being devilcat. Therefore, this one being somewhat smaller, it is is going to be called either DevilCatII, DevilCatMini, or DevilKitten, or something equally dorky. Undecided on that. Like TB2 it's mostly cnc lasercut acrylic (about 3x more than TB2) and so far it seems pretty sturdy. The drive shafts -- which have to fit the wheel bolt pattern, the bearing I.D. and the motor shaft shape -- will have to be very carefully machined, though, which officially makes this "hard". It has two 6cm omniwheel casters, which are theoretically better than swivel casters because they are stateless, but I'm a little skeptical. Of course, mounted on such short radius swingarms as they are, swivel casters were basically a non-starter. It's powered by a pair of rc-car battery packs, that is, 2x3 flat C packs, particularly 4.2AH NiMH ones, for ~60Wh total.

Some more notes on the latest:

• 0.22" instead of 0.23". (Or maybe even 0.24?)
• Confused bearing flange and body diameters.
• The motors have M3 screws. I even got to buy M2, M2.5, and M4 screws to figure that out.

I've also decided to resurrect the Logo computer system. I hadn't been planning on having a whole x86+hdd system onboard so this is starting to push the weight limits imposed by the relatively small drive motors. Fortunately the design would be pretty easily modified to accept larger side-by-side motors instead of smallish end-to-end motors, if it comes to that.

These are the main drive shafts. They're turned from 1" round 6061. One end is 1" diameter with a 0.348" radius 5 hole 4-40 bolt circle; this matches the wheels. The middle is turned to exactly 0.375" to fit a ball bearing that is set in the outer rib. The other end is drilled out to 6mm to fit the motors and has a 6-32 cross drill -- clearance hole on one side and tapped on the other -- which matches the hole in the motor shaft. These are definitely the most labor intensive part of the robot so far.

This is what it looks like so far.

This is what it looks like with the VIA N10000 mounted on top. It won't actually run until I get a wide-input-voltage psu for it, though -- I didn't really want to try a "regulated 12v" psu with the 16v battery voltage.

...that is to say, after installing the new computer psu, installing an old avr + 2x lmd18200 board I had around, giving the batteries an hour-long 75 watt jolt, it's now possible to ssh into it over wifi and drive it around. It doesn't really look different from the last photo other than some messy cables, but it *moves*. Some thoughts occur:

• I didn't actually calculate the acceleration forces when spec'ing the springs for the omni casters. They are way too wimpy and way too linear. I'm going to have to revise that part of the design with much tighter springs and zero negative deflection. (right now negative and positive deflection are both 0.5") These days it may be easier to gyro-stabilize a bot like this (like crunch) than get casters to work right.
• It doesn't really move until you give it about 60% pwm power. This is sort of annoying because it's going to be hell closing the control loops. Those beefy pittmans presently bolted into the now-defunct logo are looking pretty nice.
• There is no discernible front or back. Obviously. But now things have gotten to a point where that is annoying.

Still, I'm pretty happy with it at this point. After this many cpu hours on dc1 the atx psu had failed closed and melted the motherboard...

These are the drive motors from logo, the tragically defective and now defunct robot from which the nano-itx motherboard was also pilfered. It will take some minor frame modifications to fit these in so it may not happen for a while; also, I'm not going to bother with that until I've figured out how to build an avr servo controller capable of handling 512cpr encoders; this is sort of difficult because of the 200khz transition rate, i.e. 100 instructions per interrupt on a 20mhz avr. That's something I've been meaning to do for a while, though. The original devilcat had hardware quadrature counters, which were really quite nice, but expensive, and theoretically it should be possible to do this with a $2 avr and some clever hacking.

Also pictured are the delightfully overengineered mounts, which will not be used.

In other dc2 news, today I had new idler arms made, designed such that there is no negative deflection, i.e. the omniwheels never go below the drive wheel level. They work much, much better.