(..another) Robotic Arm Project Question

[spyrosg]

Hi, I've been thinking about 10 months now to start my first robotic project, a robotic arm. My idea is to create an arm with many DOFs (as many as possible to emulate the human arm+hand), that will be controlled remotely by a PC. I been wondering how to accomplish that and I decided to start from the arm emulation because I have some experience with graphic libraries.

So I'm thinking to start by sewing a glove () with some linear analog displacement sensors in it in order to control a virtual 3D hand in my PC.
As far as I know, Cyber Gloves are really expensive.

So, the questions are:

1) Do you have in mind where can I buy these sensors and perhaps which type specifically?
2) Is there an easy and effective way the read these sensor values from a PC in real-time?

Thank you in advance.
Cheers.

[Neophite_404]

An idea off of the top of my head might be to look at Bidirectional Flex Sensors. http://www.imagesco.com/articles/flex/sensor-pg1.html.

They are cheap and I would think an algo. to convert the change in resistance into maybe an angle measurement would be easy to come up with. (But it would take a lot of testing on your part).

That might take care of your sensing issues for things like finger joints. (If you're going that far with this). And a simple POT encoder would do for your more macro DoFs like elbows and wrists. The trouble comes in when you start accumulating all of these inputs (sensors) and outputs (servos or motors). (side note: Servos offer "embedded encoding" because you can structure your code in such a way so as to use an angle measurement to define the pulse width). I just counted on my arm and see some 27 DoF minus the metacarpals. So 2(27) for both Input and Output would be 52 I/O ports. 27 of which would likely need to be configured as software PWM and the inputs would all be ADCed in some way..... ~Insert plug for the Axon~, but I don't think even that has enough ADCs to cover everything. OOh.... I guess you could use an external ADC and then have its digital output go to a regular Gen. Purpose I/O Pin.

Hope this helps your some.

(My apologies for the rough nature of this replay.... it was sort of an off the cuff sort of thing but I wanted to get something to you quickly. I hate waiting around for replies to my posts ).

[spyrosg]

Thanks for you reply.

I was thinking of something very similar to what you suggested regarding the sensor (http://www.instructables.com/id/Fabric_bend_sensor/).

Indeed, a good model of the human hand has ~25 DoFs. Unfortunately, Axon has 16 inputs. So, I guess using more than one multiplexed ADC will do the trick. The only thing is I would not want for every 16 sensors to use a new USB of my PC.

My thoughts are to use 2 multichannel ADCs (I'm thinking of this one: http://www.analog.com/en/analog-to-digital-converters/ad-converters/ad5590/products/product.html , but for no particular reason, only because it's the only one I could find), connected to one FPGA which will transmit the data to my PC.

As you can see I'm not bothering for the output and the servos right now because it's too early and it might become too complex for me to handle at once.
Also, as you can probably see I'm a beginner at the field of electronics. I have only software skills, so I'm looking for a simple way to read the sensor values as fast and as accurately I can and take it from there. I'm guessing that when it comes to the actuators, the opposite direction will do fine for me.

Any thoughts will be appreciated.
Cheers.

[Admin]

Or buy two Axons and use a USB port for each



(mostly joking, the multiplexing method is cheaper/smaller)