What sensors to use for robot hand?

[Gramps]

This robot hand has been gathering dust on the shelf for many years.
It still needs a thumb but controlling the digits is the problem we're addressing now.
The servos were converted to continuous rotation.
We removed the pots that were inside the servos and added 5K slider pots that are positioned under the threaded rods that move the digits.
We hoped that as the digit moved, we could get the slider pots to move in unison, but it's erratic.
Using bendable variable resistors on the back of the fingers might work, but they're pricey.
Perhaps a 5 K resistor across a limit switch at each end?
Any other suggestions?
Thanks, Gramps

 

[AllyCat]

Hi,

It rather depends what you want the hand to do - just to "look" like an animated hand, or to do something "physical" such as operating a keyboard or holding an object. For the latter case I would expect some pressure-sensing to be required, even if only by monitoring the current drain of each servo motor.

To give each finger tip "touch", you might use a strain gauge, or pethaps a rubber tube "bladder" driving a barometric (air) pressure sensor such as the BMP280. To sense the "attitude" of each finger tip you might use a 3-axis accelerometer to detect the direction of gravity, but that could be quite a challenge.

Are the fingers "stiff" or do they curve, for example by using "tendons"? Are you using the additional (linear) pots to directly close the servo loops, or as a position sensor for the PICaxe ADC inputs (to indirectly control the servo motors)? For a "physical" activity, I would expect a better resolution than a servo command's 50 - 225 range might be required.

Cheers, Alan.

 

[premelec]

Perhaps the Rubbery Ruler design could be implemented with some capacity sense functions; otherwise various pressure sensitive resistance materials - as there are implemented designs commercially it would be good to research those designs...

https://www.ph.unimelb.edu.au/inventions/rubberyruler/brochure.htm

 

[Gramps]

Are the fingers "stiff" or do they curve, for example by using "tendons"? Are you using the additional (linear) pots to directly close the servo loops, or as a position sensor for the PICaxe ADC inputs (to indirectly control the servo motors)? For a "physical" activity, I would expect a better resolution than a servo command's 50 - 225 range might be required.

Yes, Allen, "the additional (linear) pots [are] to directly close the servo loops,
[and later hopefully will be used] as a position sensor for the PICaxe ADC inputs (to indirectly control the servo motors)" !
Yes the fingers curve and the servos pull the tendons
Gramps

 

[premelec]

A good way to get variable pull on 'tendons' is with PM motors & wrap of fine woven cable [e.g. like you find in an old dot matrix printer drive] - pull is directly proportionate to current in motor winding... don't try long pulls with solenoids - non linear with current and travel and mostly weak.

 

[oracacle]

For me two methods sprung to mind, personally not used either but you could look at shaft encoders and keep track of them after a homing cycle then use some kinomatics to figure out the finger position. Or use load cells on the tendons to figure out the pressure on each finger.

You could use both, but that is most likely something for further down the road.

 

[Gramps]

The homemade linear drivers we made our pretty crude. We're going to try to improve them so they are a little more linear Shaft encoders. Good idea!