View Full Version : Micro-size Dynamixels ?

10-01-2011, 02:03 PM
Hey, new member to the forum here.

I've been watching the new products coming out of Robotis for a while, and they keep pushing the envelope with more and more torque and power.
Wondering if they ever have plans to take Dynamixel in the other direction, i.e. very small intelligent servos, something that would fill the void for more dexterous manipulators.
Obviously for such products smarts would matter more than muscle, i.e. speed and power would be less important than precision.

I haven never seen the internals of an AX-12, but are there any basic reasons why it could not be downsized ?

10-02-2011, 07:26 AM
There are plenty of micro sized servo manufacturers, and a few open source replacement/upgrade control boards that when combined could fill the niche as a dynamixel would. I seriously doubt that Robotis will invest the time, energy, and funding into developing anything smaller than what they already have simply because there is not yet a terribly large demand. The robots on this board and others keep getting bigger, not smaller. Dynamixels have also become particularly popular within the humanoid classes at RoboCup, and the eventual goal of RoboCup is to make an adult human sized robot capable of competing on a full-size football* field with real humans.
There is also a problem with making actuators smaller in that eventually they cannot produce sufficient torque to support their own weight (and become ever more fragile). The folks at the DLR (http://qa.spectrum.ieee.org/automaton/robotics/humanoids/researchers-smash-super-tough-robot-arm-with-bat) and other (http://forums.trossenrobotics.com/showthread.php?3299-EX-106-RX-64-Humanoid-Robot-Giger&p=48954#post48954) places have taken the far easier, and ultimately more robust, path of placing the actuators in the wrist and linking them to the joints through high strength synthetic ligaments/tendons.

*Not that ever so annoying 'merican version where a foot almost never touches the ball. Nothing like a bunch of overweight drunks using massive RVs to punch holes into parking lots blocks away from a stadium just to watch the game on a tv and dump trash and vomit everywhere (and on everyone) within reach.

10-02-2011, 09:53 AM
Most of the cost of the ax-12 is in the electronics, so to have a servo with the same features but smaller wouldn't cost much less, if at all any less. Of course you could remove features, but there already are smaller servos out there with less features.

10-02-2011, 11:39 AM
There probably isn't enough volume to produce a single-chip solution integrating all electronics, that could be embedded straight into a small gearmotor body ?
I have worked with a few "SmartMotor" designs in industrial robotics that are basically highly integrated servos for daisy-chaining on a multidrop serial bus.

I'll see if modding a micro-sized RC servo works out for me, or i could bite a bullet and work on a custom controller for micro sized DC motor - i guess the challenge is finding a good enough feedback sensor.
Or maybe i'll just try Miga Nanomuscles out, although i have no idea how to do closed loop control of these.

10-02-2011, 10:51 PM
In the past, I needed to add 8 micro servos to a mostly Dynamixel-based robot. My solution was to create a small interface board that took in AX-bus commands on one side and output 8-hobby servo pulses. From the Dynamixel-side of things, it just looks like 8 servos. The code is checked in here: http://vanadium-ros-pkg.googlecode.com/svn/trunk/sandbox/bridge/ although it is about 1.5yrs old and doesn't support syncwrite.


10-04-2011, 06:00 PM
Thanks for that : although that obviously loses a main point of having dynamixels in the first place, i.e. position and other feedback.
I'm currently looking at hacking various sub-micro servos ( these are pretty expensive ! ) and just doing my own small feedback servos with DC gearmotors and magnetic encoders. There are some nice fully integrated small H-bridge chips with current sense circuitry already inside.