Remove this ad

Lead

Jan 25 16 5:55 PM

Tags : :

PROBLEM
The Ada hand can fail to grip objects or grip objects too hard or too softly to pick them up properly.

CHALLENGE
To find a way to make the Ada hand more adaptable to gripping objects with varying force.

REQUIREMENTS
1. The solution must enable the Ada hand to grasp objects with an appropriate strength
  • See problem
2. The solution must be firmware based
  • Hardware can be used in combination with firmware but a hardware only solution is already being explored here
3. The solution must be licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
  • Since the original Ada hand is released under the this license, any derivatives must be distributed under the same license.
SUGGESTED BRIEF
Develop a force sensing algorithm for the hand which detects how much force is being applied to the fingers and reacts accordingly. When being used for robotics applications this functionality could enable robots to grip objects with a pre-determined force, and increase their grip automatically if they detect an object is slipping. The PQ12 linear actuators have built in positional feedback, which when measured over time and integrated, can give an estimation of the speed at which the fingers are closing. This can then be compared to the speed expected if no object is in the way of the fingers, so if they're moving slower than usual then there is an object obstructing them. The slower the fingers are moving in relation to their programmed speed, the harder the grip. Since the PQ12s have a bit of backlash in the servo horn it is possible to detect small movements of the fingers even when the actuators aren't moving. An alternative option might be to use force sensitive resistors in the fingertips to sense pressure applied by each finger.

FURTHER SUGGESTIONS
1. Build the solution into fingerlib.h so it's easier to track the changes.
 
Quote    Reply   
Remove this ad
Remove this ad

#1 [url]

Jan 27 16 7:19 PM

Hi Joel, do you have a UK supplier for PQ12? Preferably cheap :) Or is there an alternative part I can hack with? I assume a simple tunable PID controller is what is required, taking in the target position and measured position and updating accordingly.
Cheers jd

Quote    Reply   

#2 [url]

Jan 27 16 7:56 PM

Using a transducer on the finger tips to measure the amount of "slippage" and force being applied would be a good starting point to start working on either a control loop such as PID, or an Artificial Intelligence approach such as Fuzzy Logic.

With PID you could tune it to apply more or less force depending on the amount of slippage.

Using Fuzzy Logic (Mamdani or TSK) you could output a fuzzy set (or a linear function for TSK) which states the amount of force that would need to be applied depending on the inputs into the system, that could be the data from the slippage / force applied sensors.

Quote    Reply   

#3 [url]

Jan 28 16 10:24 AM

Hi John,

We supply PQ12s! Free delivery in the UK (and 100% of profits go back into poduct development).

Robotshop also deliver to the UK.

I'm afraid they just aren't cheap. If we end up getting enough sales we'll be able to buy them in larger volumes and will sell them cheaper. The gearmotor inside the firgellis is a lot like this one (£8 from RS). We have made a little prototype of our own linear actuator with FDM parts and a standard M3 bolt and nut as the leadscrew but it's not likely to match the firgelli (maybe with SLS parts). Could be something for you to look into ;) 

Quote    Reply   

#4 [url]

Jan 28 16 10:27 AM

John wrote:
Using a transducer on the finger tips to measure the amount of "slippage" and force being applied would be a good starting point to start working on either a control loop such as PID, or an Artificial Intelligence approach such as Fuzzy Logic.

With PID you could tune it to apply more or less force depending on the amount of slippage.

Using Fuzzy Logic (Mamdani or TSK) you could output a fuzzy set (or a linear function for TSK) which states the amount of force that would need to be applied depending on the inputs into the system, that could be the data from the slippage / force applied sensors.

Hi John,

Nice idea, would the ATMEGA2560 be capable of a fuzzy control algorithm? (not my area of expertise) What kind of transducer would you be thinking of, peizo-electric or resistive force or something else?

Quote    Reply   

#5 [url]

Jan 28 16 12:12 PM

JoelGibbard wrote:
Hi John,

We supply PQ12s! Free delivery in the UK (and 100% of profits go back into poduct development).

Robotshop also deliver to the UK.

I'm afraid they just aren't cheap. If we end up getting enough sales we'll be able to buy them in larger volumes and will sell them cheaper. The gearmotor inside the firgellis is a lot like this one (£8 from RS). We have made a little prototype of our own linear actuator with FDM parts and a standard M3 bolt and nut as the leadscrew but it's not likely to match the firgelli (maybe with SLS parts). Could be something for you to look into ;) 

I'll get a PQ12 kit off you in a few days, and an almond board :) For the Almond board, I assume it doesn't need a regulated input? I am hoping to run it off 3S Lipo. Have you considered selling consumables like filament? It could be a way of getting regular, though perhaps low margin, cash flowing. 
Cheers jd

Quote    Reply   

#6 [url]

Jan 29 16 10:25 AM

jaundice wrote:

JoelGibbard wrote:
Hi John,

We supply PQ12s! Free delivery in the UK (and 100% of profits go back into poduct development).

Robotshop also deliver to the UK.

I'm afraid they just aren't cheap. If we end up getting enough sales we'll be able to buy them in larger volumes and will sell them cheaper. The gearmotor inside the firgellis is a lot like this one (£8 from RS). We have made a little prototype of our own linear actuator with FDM parts and a standard M3 bolt and nut as the leadscrew but it's not likely to match the firgelli (maybe with SLS parts). Could be something for you to look into ;) 

I'll get a PQ12 kit off you in a few days, and an almond board :) For the Almond board, I assume it doesn't need a regulated input? I am hoping to run it off 3S Lipo. Have you considered selling consumables like filament? It could be a way of getting regular, though perhaps low margin, cash flowing. 
Cheers jd

Great, thanks John, I'll throw in a broken firgelli as well so you don't have to take a working one apart in case you wanted to see how they work to look into making your own.

The Almond board doesn't need a regulated input, there is a 5v regulated supply on the board but the firgelli motors are driven directly by the main supply voltage. Hence you can run it at 12V with 12V rated firgellis (as in our instructions) or you could buy 6V firgellis and run it at 6V and it will work just the same, the only difference being that the 6V motors would draw twice as much current. It will run fine with a 3S Lipo, that's how we generally test it.

We have thought about selling consumables. Filament is a pretty comititive market though, there's no way we could comete with 3D Filaprint in the UK. We'll keep thinking on it, thanks for the tip ;)

Quote    Reply   

#8 [url]

Feb 15 16 11:06 AM

I don't see why you couldn't use an  ATMEGA2560 for a fuzzy control system. Engineering / electrical engineering aren't my strong points, but from a little bit of knowledge in transducers I would probably use resistive. 

I think the good thing about this idea is you won't need to know what sort of material you are holding. If you were working out the coefficient for friction, you would have to know what surface you are on but using some feedback loop / fuzzy you are rather looking at having 0 slippage, so a small - higher pressure can be applied depending on this error rate.

Quote    Reply   

#9 [url]

Mar 3 16 3:32 AM

Hey, have you tried implementing current sensing into the controller? There was some work done a while ago at: http://openservo.com/ on replacing the simple servo controls in a hobby servo with a full PID closed loop control over I2C. A similar idea might be very beneficial to develop better logic for the hand.

Quote    Reply   

#10 [url]

Apr 1 16 9:01 AM

jimshealy wrote:
Hey, have you tried implementing current sensing into the controller? There was some work done a while ago at: http://openservo.com/ on replacing the simple servo controls in a hobby servo with a full PID closed loop control over I2C. A similar idea might be very beneficial to develop better logic for the hand.

Good idea Jim,

We're doing this in the next version of our electronics.

Quote    Reply   
Remove this ad

#11 [url]

Jun 19 16 9:19 PM

varying grip strength

Instead of focusing on using sensors for the grip strenght, as these are available in the tips of fingers in todays prosthetics but are extremely expensive. Instead we could try focusing more on the available cheap sensors and perhaps program it to read the flex in the muscle of the forearm. for example the harder you squeeze your muscle the stronger the grip in the prosthetic hand would be, yes this would take a little practice but i dont think it would be too complicated, it would keep the cost down, and would still get the varying grip strength. Please send feedback on thoughts.

Quote    Reply   

#12 [url]

Jun 19 16 10:57 PM

jbrawley wrote:
Instead of focusing on using sensors for the grip strenght, as these are available in the tips of fingers in todays prosthetics but are extremely expensive. Instead we could try focusing more on the available cheap sensors and perhaps program it to read the flex in the muscle of the forearm. for example the harder you squeeze your muscle the stronger the grip in the prosthetic hand would be, yes this would take a little practice but i dont think it would be too complicated, it would keep the cost down, and would still get the varying grip strength. Please send feedback on thoughts.

We do indeed do this in the latest version of the software. It is also implimented on traditional myoelectric upper limb prosthetics.

The advantage of fingertip sensors would be that there would be potential to offer some feedback to the user, so they know how firm their grip is on something without necessarily looking at it. We've tried this in a very rudimentary test and it seemed to work quite well.

Quote    Reply   

#13 [url]

Jun 19 16 11:05 PM

o I see, Have you found sensors that can be placed in the finger tips that can be purchased for reasonable price? if so would you share the sensors you are using?

Quote    Reply   

#15 [url]

Aug 7 16 5:55 AM

Force Sensors and ways of sensing them

Polilu have a variety of force sensors that could fit your needs.

https://www.pololu.com/product/1695

A lot of robotic supplies and electronic shops have them.  Also re the feedback - why no embed some of the tiny phone vibration motors under the shield so that the wearer can feel them. The tighter the pressure - the greater the vibration.

It's cheap and an Arduino pro would be perfect to handle this one function since the feedback mechanism is the user themselves.

Sincerly
Marc
 

Quote    Reply   

#16 [url]

Aug 7 16 5:58 AM

Vibration motors for sensor feedback

A very cheap way for a user to sense feedback would be to use the micro motors used in Mobile phones to vibrate them. Using Pwm you can control the speed of vibration and so give feedback from a soft touch to a firm grip

You can get this tiny motors from many electronic hobby shops and again - they are cheap, low powered, and could easily be powered from an Srduino and a ULN2003.

Sincerely
Marc

Quote    Reply   

#18 [url]

Aug 21 16 11:54 PM

would electrical conductivity differential help

Perhaps it could be a linear potentiometer. I think the person who said to think of the hand ad not straight. I am curious if it also closes inwards kind of making a fist. Is this caused by resistance caused by the contact of the grabbed object. Also I am wondering if we make it like a multimeter can it sense a difference in electrical conductivity when it comes in contact with the object.

Quote    Reply   

#19 [url]

Nov 14 16 11:41 AM

Hi,
Do you think an algorithm that measures the coefficient of friction of the surface the object is placed and then it is used to find the frictional force acting on the object when it is tried to move. So that the same amount of the force input can be given to the finger tips to hold the object.

Quote    Reply   

#20 [url]

Nov 16 16 9:02 AM

sai7273 wrote:
Hi,
Do you think an algorithm that measures the coefficient of friction of the surface the object is placed and then it is used to find the frictional force acting on the object when it is tried to move. So that the same amount of the force input can be given to the finger tips to hold the object.

I'm not too sure this would work as you would need to know what object you're currently holding. For example, ice would have a much less lower friction value then say rubber. So, without knowing what object you're holding I don't think this would be possible. 

Quote    Reply   
Remove this ad
Add Reply

Quick Reply

bbcode help