Date: Wed, 15 Mar 2006 01:48:06 -0600 (CST)
Subject: first successful driving test
X-UID: 149

The media is uploaded to the website as it is a large update.

http://golem5.org/robot1/images/img2809.jpg
     -- laptop and robot with a few people watching

http://golem5.org/robot1/images/img2811.jpg
     -- front wheels stopped by skid plate

http://golem5.org/robot1/images/img2814.jpg
     -- wheels almost ripped dangling wires off

http://golem5.org/robot1/images/img2816.jpg
     -- nylon cord came off bobbin so wheel is stuck

http://golem5.org/robot1/video/mvi2810.mpg
     -- confusion while turning camera mode dial

http://golem5.org/robot1/video/mvi2812.mpg
     -- robot is 60 to 70 feet away, trying to drive it back

http://golem5.org/robot1/video/nightrun01.mpg
     -- view from robot camera while trying to drive out of an alley

The good:

1. The pinned square nuts for the rear wheels held up. While the peak
    torque of the motors should be able to apply a shearing force about
    three times the failure point of the dowel pins, this doesn't happen in
    practice. Heat treated and stainless steel are quite strong.

2. The nylon cable and bungie cord steering system works when there is
    enough tension. The wheels have excellent range of motion with
    sufficient turning torque.

3. The network camera provides a decent image at a good framerate.

The bad:

1. The steering motor arrangement is still a difficult system. I
    anticipate feedback control in software. However, most people find it
    very strange and suggest switching to a large servo.

2. The exposed wiring needs to be covered by a shell for protection. The
    wires from the steering motor and cooling fan battery pack were nearly
    ripped off the terminal block by the right front wheel.

3. The cable system in front needs improvement. Cables can come off the
    bobbin. When this happens, the wheels become stuck.

4. The skid plate has protruding nuts and threaded rod that can become
    stuck on the ground?

I learned quite a bit tonight.

The robot is essentially undriveable without software control. An
experienced RC driver might be able to do it. Even a rudimentary feedback
control system is required. I had a lot of difficulty driving the robot
out of a wide alleyway.

Night and darkness offer probably the best conditions for testing.

* electronics remain cool
* structured light vision might actually work
* laptop LCD is visible
* almost no traffic or police in the Garland warehouse district

I'm cross-compiling MICO, a free CORBA implementation. It will require
some hacking to get it built under uClibC/buildroot. The gcc in the
buildroot toolchain lacks exception handling (reasonable choice for an
embedded system) which is strange for CORBA as it throws exceptions
everywhere.