Subject:
|
Re: Extra RAM
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Fri, 23 Jun 2000 12:08:12 GMT
|
Original-From:
|
Andy Gombos <gombos@ne*SayNoToSpam*.infi.net>
|
Viewed:
|
676 times
|
| |
|
|
To find doors, you can assume that they are are always on a wall. Then, to find an
open door, use an edge finding algorithm to highlight the edges. Where there should
be a wall, but is not, there must be a door. A closed door could be found by looking
for areas of wall that are not in the same plane as the rest of the wall. This would
not work on flush doors though, that are even with the wall.
Andy
Carsten Müller wrote:
> John Barnes wrote:
> > Ah! I think I see the difference. My concept of the "map" is something that
> > starts out blank and the prime objective is to fill it in.
>
> If you start with a blank map, you don't need a full sized array for
> mapping. You can build up a linked list or a binary Tree where each node
> hold position information status and size. If your memory is nearly use
> up. You start a garbage collector which tries to find rectangular areas
> of the same status, which are than replaced by one Node which holds this
> information. If your garbage collector is smart enough, it may also find
> areas which are not visited and which can't be visited by your explorer
> at all. The software for such sparse array technic is a little bit more
> complicated that using a real array but the memory advantage is really
> great, if you have big areas of the same status, which should be normal
> for a high resolution map.
>
> > I suspect the
> > use of a term like "door" implies a map supplied from outside. In which case
> > I agree that an object based expression of the surroundings is far better than
> > a raster based one. Unless .. you have an algorithm which allows your robot
> > to determine that something is a door! In which case I want to know a lot
> > more about how you do that!
>
> The answers are Yes and No. I start with a generated map and there is no
> algorithm for detecting a 'door' yet :)
> Till now, i only have code for handling the spatial information and it
> is not completed yet. My actual work is focused on positioning in a
> controlled environment using a combination of dead reckoning and
> triangulation using fixed IR beacons.
>
> Regards Carsten
|
|
Message is in Reply To:
 | | Re: Extra RAM
|
| (...) If you start with a blank map, you don't need a full sized array for mapping. You can build up a linked list or a binary Tree where each node hold position information status and size. If your memory is nearly use up. You start a garbage (...) (24 years ago, 23-Jun-00, to lugnet.robotics)
|
4 Messages in This Thread:
  
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
