Build a $100 Robotic Arm with CCD Motion Capture

[tusavision]

http://builders.reprap.org/2009/09/bambot-bamboo-robot.html

I want to see this design move forward. Before I can work on the mechanical engineering aspects of the motors: I want to work on my position control system.

(Cheap geared DC motors like those from solarbotics/chinese drill motors seem ideal if I can convert them into servos somehow.

I'm also considering DIY worm wheels drawing from the astronomy or http://en.wikipedia.org/wiki/Magnetorheological_fluid between clutch plates to manage the holding torque issue with DC, but that's getting ahead of myself.)

Rather than going with the potentiometer/optical encoder positioning that seems popular on many systems/servos: for ultra-low budget reasons: I'm inclided to go with a CCD active motion capture setup wiimote style.

http://www.chinavasion.com/product_info.php/pName/micro-wired-pinhole-color-audio-camera/

Now: If I put and IR filter on several of these cameras and affix them to the arm on each axis of movement: I imagine I can keep the robotic arm in a room with line of sight to multiple reference IR LEDs, and I can use this code:
http://code.google.com/p/lasertraq/

to translate the video feed into positional feedback on the arm.

I guess is what I'm looking for is suggestions/constructive criticism. Trouble you anticipate I might run into from your experience in such things. Not being a CS/Mechatronics expert, I have no experience to draw from on the: "in over my head" scale.

Flame away.

 

[tusavision]

Talked to several people about this. One thought I should use this motion capture technique to supplement traditional positioning feedback from an optical encoder. The other didn't have an opinion one way or the other but didn't see anything immediately wrong with the idea. Some discussion about the possibility of lag.

The "geckodrive" would be one option for controlling it, although: these solarbotics motors are so low voltage: that may be overkill.

It didn't stike me at first, but the idea of using bamboo as a material really tickels me. Competitive with acrylic, however universally available, and less brittle.

Can anyone offer a prediction on lag/anticipate any problems with this approach before I start sinking money into it?

 

[tusavision]

Perhaps this is the wrong subforum? Can a moderator move me to Electical engineering or computer science? Something more relavent? I think the real issue at this point is an estimation of the lag that such a control system would have to deal with.

 

[mgb_phys]

I think the ABB flexpicker robot uses image sensors for positioning - but with a traditional jointed arm it's going to be easier to just use encoders or encoders+inclinometers on the arm segments

 

[Solveer]

I think you will need to consider what is available and easy enough for you to do. If you have serious equipment at your disposal and sufficient time, it would certainly be nice to work out a more advanced system. However having said that, you may want to consider starting at the low/easy end and then working your way up. From my experience with robotics/motion control (designed & built 4 different CNC routers/mills & 2 robotic arms), you may want to start off with simple RC servos. They are easy to control and take care of themselves. A step up from that would be a system with stepper motors & perhaps encoders (if you don't want to run this in open loop). The optical system sounds like overkill. The issue most likely with that would be accuracy/resolution but also complexity.
But good luck anyway.