Subject:
|
Re: A robot who knows his position (fwd)
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Thu, 29 Apr 1999 07:14:29 GMT
|
Viewed:
|
1011 times
|
| |
|
|
In lugnet.robotics, lego-robotics@crynwr.com (John A. Tamplin) writes:
> I have been thinking about a different approach. I have not actually built
> anything to use it yet, but I believe it should work reasonably well.
>
> The basic idea is similar to the old optical mouse pads -- draw
> horizontal and vertical lines on the playing field. Use different colors
> for the horizontal and vertical lines (easily distinguishable by the
> light sensor's greyscale values). You can then compute relative position
> changes and from that direction of travel. Obviously, the resolution you can
> achieve is limited to the spacing of the lines and the accuracy of
> measuring your bot's motion in between them, but it should be sufficient.
> The difficulties come from distinguishing the lighter-colored line from the
> fringe of the darker line, and the varying line widths depending on the
> angle of travel relative to the line. For instance, travelling parallel
> to one of the axes will result in no line crossings for that axis (or
> continuously being on an axis).
>
> You can use wider lines in the center of the playing field to give an
> idea of absolute positon, or you can allow your bot to explore the
> playing field determining the dimensions by bumper contact with the edge.
>
> Clearly, this is only applicable in limited situations where you can mark
> the area the bot will traverse.
I like John's idea. You could combine it with odometry and get a very
affordable and precise positioning system. I'm quite happy with the results of
odometry in the short range, I mean until you accumulate too many small
errors. My robot estimates its position very well (much less then 1cm=0.4")
when the run is short (about 1m=40"). If you give it a way to zero the
accumulated errors using an external reference (like John's grid) you can keep
this resolution constant during time. The grid could have a step of about 10"
or even 20" if you rely on odometry between the lines.
Mario
|
|
Message is in Reply To:
 | | Re: A robot who knows his position (fwd)
|
| (...) I have been thinking about a different approach. I have not actually built anything to use it yet, but I believe it should work reasonably well. The basic idea is similar to the old optical mouse pads -- draw horizontal and vertical lines on (...) (25 years ago, 28-Apr-99, to lugnet.robotics)
|
14 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
