Subject:
|
Fun with Differentials
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Wed, 25 Nov 1998 16:29:59 GMT
|
Viewed:
|
2173 times
|
| |
|
|
That differential brick really is a godsend-
I built a Lego robot for my AI degree in 1995, at the University of Nottingham.
This had a Lego chassis, and one of MIT's WAC-bot circuitboards as a
controller. Software could be written and compiled into the board's native
assembler language, and then run in the robot autonomously. I think this board
must have been one of the early forerunners of the intelligent brick and the
Mindstorms system.
One of the themes of the project was employing a homebrew polarised-light
sensor to gauge orientation, in the manner of honey bees. For this I needed to
provide my robot with 2 modes of movement- rotary (where the wheels turn in
opposite directions) and forward/backward (same direction). This was so that
the sensor, mounted above the middle of the main axle, could gather data
through a 360 degree spin without moving off-centre. Unfortunately the board
could only run 1 direct-drive motor (like the Lego Mindstorms ones) and one
motor with a limited range of movement in both directions (it couldn't rotate
completely).
Solution: a Lego gearbox, based around the differential brick, which could
switch between these modes. Power to both wheels was provided by the single
direct drive motor, and there was a sliding clutch to engage/disengage the
differential, which was operated by the limited motor.
This arrangement worked very well, and I have some high-quality close-up video
of the mechanism. If I can digitize this I'll post a link to it some time. The
main problems I encountered were meshing the teeth of Lego cogs successfully.
There's not really enough precision with Lego for reliable meshing, but it is
possible. The 'curved-tooth' cogs help, but I found I could never guarantee
meshing every time.
In order to get around this, I incorporated a switch sensor which operated as a
proprioceptor (a sensory mechanism which measures some internal state of a
robot/organism). The switch would only be depressed if the clutch had moved
across fully, and had therefore meshed successfully. If not, the robot knew to
retract the clutch and make another attempt at engaging it! Again, this worked
nicely, and the clutch almost always engaged after a maximum of 3 attempts.
I must get around to rebuilding this, with the new kit, and perhaps a better
clutch. I'll dust off my project file, and see if there are any other relevant
robotic tales lurking inside.
Keep posting!
Mat.
[ Currently at the University of Sussex, doing MSc in Evolutionary and Adaptive
systems. Also working as web-developer for London Web Communications
http://corp.londonweb.net ]
|
|
1 Message in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
