Lazy Cleaner 9000

Color diff. is now implemented and give a nice easy numb of how close 2 "objects" are in color. No more RGB rough approximation.

Tonight I will get the code on GITHUB so anyone can look at the rough code. It needs lots of tweaking (format, comments, faster routines, etc....).

LICENSING:

OFFICIAL 2 MIN VIDEO:

HIGH LEVEL DOC:
https://github.com/TheOriginalBDM/Lazy-Cleaner-9000/blob/master/docs/clean_sweep_diagrams.pdf

Another quick search on Inverse Kinematics for Raspberry Pi turned up a project for the meArm that is exactly what I was thinking about. I have forked a copy to my account in case I need to make changes. Once this is implemented I can put X,Y,Z (fake Z) coordinates on the trash and have the arm navigate to the specified location. Once I have this implemented it will complete Phase 1 of the project in being able to detect trash and pick up. The next milestone will be getting the robot to move to trash out of reach, pick up, and navigate to next piece until it looks clean.

https://github.com/RorschachUK/meArmPi

Big dog and I were finally able to get into the garage and finish building the arm. Went smoother than I thought even though it was at least 100 degrees in the garage. Enjoy the pics. Yes big dog left me 10 minutes into building to do his own thing.

I did a little research yesterday in how to figure out the best way to have the robot pick up objects and place into "the mouth". This weekend @ night I'll look more into this but I may need to install PyBrain onto the Pi to assist with this. It seems like one way to to this. I'll also post more details on the mouth. I have some roomba wheels which will feed pull objects in :).

I am new to robotics to I have a theory (which is probably wrong). Do I really need depth? Can it be simulated with a single camera? Is everything just a co-ordinate since all distances are limited by the arms (move arm to approximate coordinate, distance measure to object on arm, finish grasp)?

just in case I'm keeping these links safe for later.

Kinect stuff:

http://openkinect.org/wiki/Main_Page

Machine Learning Links:

http://pybrain.org/

http://deeplearning.net/

http://www.cs.ubc.ca/research/flann/

Tomorrow I will add the elbow and hopefully by Friday I will have a shoulder working. That would complete the arm's since the other side is just built in reverse.

Family said change the name back to my lazy 9000 series. Done! Moving on I sort of mapped out the flow of data and much is to be done. I am looking for an python actor framework to use since so many moving parts and checks are involved. For now I am going to use gevent to keep moving ahead. The servo controller is out for delivery today as well.

Builta quick prototype of what the claw will be to grasp objects. Once I can have a demo working I will move to aluminum. PWM code coming this weekend.

Uploaded a few pics for Mike but the garbage eating beast is still in hiding. Just think of a crab w/o legs and other features, that is my proto bot. Images shows easy identification of objects of non-earth tones with area as the number presented for each rectangle. Per wife's suggestions if this does not make it to the next round it will be a pooper scooper for our 3 dogs that crap about 50lbs (seems like) a day in the backyard.

Hardware has arrived. Time to upload what I have worked on and see what happens.