PDA

View Full Version : Gearing up the AX-12



ohbowz
10-04-2011, 12:30 AM
Hello, I've been working on these Robot legs for awhile, I have some motion tests:


http://www.youtube.com/watch?v=8ypH1asPqtQ

At this point I feel it isn't very precise in its motion, it has some wobbles at times, and the resolution is too low for the length of leg (around 2ft from hit to toe)
I am playing around with gearing it up 2:1 to get more resolution, I assume at the cost of speed.
I have never done anything like this before, here is what I was thinking of doing.


http://blog.ohbowz.com/wp-content/uploads/2011/10/gearup1.jpg

http://blog.ohbowz.com/wp-content/uploads/2011/10/gearup2.jpg

Has anyone done this to the AX-12 servos? Any advice for someone who hasn't done this before?
The gears are 48 Pitch, 60 tooth to 120 tooth.
I also realize I may need to have the servos set to continuous rotation. I understand this is possible with these dynamixel servos.. right?

Thanks for any help!

-joe

iBot
10-04-2011, 04:33 AM
The continuous rotation mode does not give 360 degree accurate positioning. In continuous rotation you set speed not position, and the internal angle sensor only covers 300 degrees.

The good news is that for most joints a 150 degree movement is sufficient. For joints like hip pitch which do require larger angles, high resolution, and higher torque, then maybe better to go to MX-28.

ohbowz
10-05-2011, 03:48 PM
The mx-28 is pretty pricey, I am trying to stay within a budget. If I double up the ax-12, would that increase the torque?



Thanks for the info on continuous rotation.

cire
10-05-2011, 05:48 PM
If you double up the ax-12, it will almost double the torque available (but it will also weigh twice as much).

I really like the concept of gearing up a ax-12 for a joint that does not need the full 300 degrees of rotation - such as a ankle, I consider doing this for joints like that (or using a linkage to effectively gear the joint).

As for your resolution problem, the gearing would help, and shortening your legs would help a lot. This also would help with the torque required at the joints.

Slugman
10-05-2011, 08:51 PM
@ cire, that's exactly what I did for the ankles - Still sorting out my CAD drawings so I can get the parts made properly, but for ankle tilt, I found a lever 4mm from the servo horn centre pushing down on the foot with the tilt hinge set 22mm away from the servo centre was enough tilting movement for a typical AX-12 biped to walk on a flat surface, with a tiny bit of leeway. I made 4 prototypes so I could do the hips as well. Unfortunately the prototypes are very heavy. :sad:
@ ohbowz, those legs are very long. I doubt you would need more than a few degrees tilt to move the COG of your robot over one foot, but you will probably have hassles with slop in the joints & servos. I'm not an expert by any means, but I agree with cire about shortening the legs. It would make it a lot easier. My bot with legs about 20cm from foot plate to top of hip had a lot of slop, which became worse with the levers, & I was trying to keep the tolerances pretty tight between the bolts & the lever/hinge. Try comparing the bot in the standing & stepping positions & work out how much servo rotation you need to get the required ankle tilt with the big cog. If you can keep the cogs instead of using levers, & use the normal range of servo motion (I.E. Not 360 deg rotation) to retain the servo feedback then I would do so, for various reasons. It's easier with feedback, for starters... :happy: