Subject:
|
Re: a couple of new-bot questions from a generally quiet lurker
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Tue, 21 Sep 1999 00:22:31 GMT
|
Viewed:
|
618 times
|
| |
|
|
In lugnet.robotics, lego-robotics@crynwr.com (dave madden) writes:
> I think the best idea I've heard for the differtrans is to build a
> regular, one-motor-per-wheel platform, but also drive the differtrans
> inputs from the wheel axles. Then, use a rotation sensor on the
> differtrans outputs to see whether the robot is turning due to
> different axle speeds. This is a win because you don't have to
> transmit any power through the (lossy) differtrans, you get both
> motors driving the robot, and you still know whether your robot is
> going straight. (I didn't think of this; somebody else on the list
> did. It was a while ago, but maybe you could find the message in the
> archives. Or maybe the originator will speak up and tell us how
> things worked out for that robot...)
I assume that was me; I posted a message with a link to an LDraw file of a
robot I came up with that uses that design. I can't take all the credit for
it, however: I was inspired by Simen Svale Skogsrud's web page on his Vector
Rover using fiber-optic rotation tracking, and I expanded on his idea to come
up with my design. If you'd like to see my original threads, run searches in
lugnet.robotics and lugnet.cad.dat.models for *rover* and *transmission*.
Otherwise, here are the indicated posts:
1. Robot base, RCX, and wheels
http://www.lugnet.com/cad/dat/models/?n=210
2. Differtrans-based decoder, to mount on the back of the robot base and
couple the wheel motion to the rotation sensors
http://www.lugnet.com/cad/dat/models/?n=214
One caution: the transmission post has the motor output GEARED UP to go to the
differentials; I've had some problems with the robot getting confused as to
rotation the way it's indicated. The rotation sensors can handle direct motor
output at speed 8, but not much more than that. Change the 8t and 40t gears
on each side to 24t gears, and it should work fine.
I also managed to snag two rotation sensors which I used to calculate
rotation, which are more accurate than the fiber-optic units. The robot as
posted works very well for me; since it tracks both forward/backward and
left/right movement, it can easily move a specific direction in a plane. This
allowed me to develop functions for movement: Fwd(12) and it would move 12
inches forward, counting rotation sensor ticks; Left(45) and it would rotate
left 45 degrees.
In addition, I'm happy to say that it meets all the criteria of another post
on this subject:
> 1) Like the single rotation sensor system, you can monitor the
> difference in wheels in forward/backward movement
> 2) You can monitor Forward/backward DISTANCE
> 3) You can monitor turning distance.
> 4) You can detect stalling. (if no pulses are coming in, the robot is
> not moving, and something else must be done.)
> 5) You can calculate speed.
The only thing I don't like about this design is having to correct "drift"
every few feet, although with this design you know about it and can fix it
easily on the run. Other than that, it works pretty well.
|
|
Message has 1 Reply:
Message is in Reply To:
9 Messages in This Thread:
  
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
