Date: Fri, 3 Feb 2006 02:52:07 -0600 (CST)
Subject: robot hardware done
X-UID: 131
Content-Type: IMAGE/JPEG; name="img2584.jpg"
Content-Type: IMAGE/JPEG; name="img2585.jpg"
Content-Type: IMAGE/JPEG; name="img2586.jpg"
Content-Type: IMAGE/JPEG; name="02-19800521081842-00.jpg"

I'll have to perform a final system test either later today after sleep 
and work or on Saturday. I'm confident everything is wired correctly. I 
was careful. I know the onboard computers are ok as I left them running 
for about two days. The first day, they were running the V4L application 
motion - turning the robot into a giant network camera. Attached is a 
frame out of the robot's right camera of a maintenance guy who entered my 
apartment. I forgot to set the system date so the timestamp is wrong.

Believe it or not but I'm on schedule. The plan was:

one year to learn basic electronics and embedded systems principles
one year to design and build the robot hardware
one year to develop the robot software

The first two stages are done. Now is the last stage and really, the most 
complexity. I have to re-learn much mathematics. I no longer feel 
comfortable with matrix computations and Hilbert space thinking. Signal 
processing depends on both of these. I don't know what a Kalman filter is 
and I've forgotten how the FFT (fast Fourier transform) works. Conjugate 
gradients, Lanczos methods, etc are all faded out of memory. On top of 
this comes a lot of control theory and signal processing I was never 
exposed to in university. Crunch time!

Additionally, I believe that open source software is the best path for 
robotics and AI. There needs to be a general system architecture and 
modular software model for robotics systems. If this exists, then 
development can be distributed. Today, robots are each unique from 
hardware to software. To a large extent, everyone starts building from 
scratch each time. This is bad model for development.