Prototype zero-backlash joint with muscle-inspired actuators and control

Hi all, I’ve got this prototype of a zero-backlash joint made with very cheap components - DC motors, potentiometers, rubber bands, pieces of string, an arduino for control - and with a magnetic angle sensor in closed loop, it can do very precise angular position control, with some 0.05 degrees of steady-state error.

Does this sound like it could be useful in microscopy or it should be more precise? Potentially, the muscles could be rearranged to move things in 2D, and also potentially things could be made more precise with a higher resolution sensor (I’m using a 14bit ams sensor on the joint axle)


On the left are the two motors. They are twisting strings and pulling on tendons. There are two linear sliding potentiometers (in the white area, right of the motors) that measure the lengths of twisted strings.

The tendons are, well, just pieces of string, but they don’t twist. They each have a tension sensor - the black rectangular thingies. The tension (or force) sensor measures the stretch of the rubber band using a very small (1cm travel) sliding potentiometer.

Finally, on the right is the joint. When the upper force is higher, it moves CCW, when the lower force is higher it moves CW.

It can also stay in the same angle, with both ‘muscles’ pulling more strongly, as in isometric contraction in human movement.

As you can probably see from the image, I’m no mechanical engineer, so these things could probably be improved.

More details in this extended abstract from a conference: [pdf]