Pinchy the robot arm is doing well, and is way fast thanks to two (well, three) rounds of motor upgrades. She’s now outfitted with six GM8 gearboxes, each with an RM2 high power replacement motor. Standard servo electronics weren’t able to keep up with the higher motor current, so Pinchy’s motors are now driven by a mix of quarter-scale servo control boards with big discrete FETs (Hitec HS700, HS805, and Futaba S3801, happened to be what I could find lying around).
The next problem I want to tackle is positioning precision. In their former lives running hobby servos, these controllers probably had a resolution of about 1.5 degrees, and I’m not getting results nearly that good now. These controllers are nice and simple to use, but they won’t deliver much power in response to very small position errors. What I really need to do is set up my own microcontroller to measure the joint potentiometers and my own motor controllers so I can tighten up the control loops. Or, does anyone know a supplier of cheap ~1000 tick quadrature encoders?
I still don’t quite regret getting rid of the original stepper motors, but potentiometers and DC motors aren’t turning out to be the perfect solution I originally thought they would be. Although, trying to account for the cable redundancy in software would also have driven me crazy. Sometimes it’s really hard to compare hardware/software solutions to a problem, especially when they’re both convoluted.
I was really hoping to get Pinchy playing chess for a group final project in a robot arm theory class I’m taking. We had managed to convince ourselves that we could overcome the Pinchy’s imprecision by always approaching pieces from the same side, only making large fast motions, etc… As of this morning, however, Pinchy has had her chess-playing privileges indefinitely revoked. I thought I had the pawn-pickup trajectory just right when, well, this happend:
It’s one of those things that’s just statistically inevitable. If you play with robot arms, you will eventually smash them into something delicate. I haven’t had chicken pocks or crashed a car, so I guess I was due. Amazingly the touch-pads all still work, and the arm itself is fine, but it’s time to focus on the trajectory planner itself. Making an actual robot arm move is not within the scope of the class, and thus won’t improve our project grade (LAME!). Pinchy will eventually do something neat, but at least for now, playing chess isn’t it.
-Adam