Subject:
|
Re: Has anyone been experimenting with mapping algorithms?
|
Newsgroups:
|
lugnet.robotics.handyboard
|
Date:
|
Sun, 29 Nov 1998 04:06:26 GMT
|
Original-From:
|
krasing@^nospam^iastate.edu
|
Viewed:
|
2594 times
|
| |
|
|
For a class project, I found a few alternatives for obstacle avoidance
with known areas using evolvable code. You could call it mapping, I
suppose, though it maps for a specific type of problems - those that have
a "goal" of some sort (I used a lightbulb and analog photoelectric
sensors for the light detection). I'm in the process of translating i
the best method (neural nets) over to IC (it's currently in C++). I do
not remember how big it was in C++ and as I'm using an older version of
IC, I do not know how big it will be.
Basically what it does is use a digital touch sensor to mark the
obstacles (walls, etc.), and as it evolves, it learns to avoid these
things. By using random "goal" and "start" positions in the evolution of
the neural nets, it learns to avoid the obstacles and when in use, it
does not need a "goal" after the net is evolved. I don't know if this is
what you were talking about or not. It's fun to play with, though. :)
On the 350MHz network I used to run the code for a known, simple course
(four walls, two walls offor "obstacles"), it took maybe 2 minutes to run a
few thousand neural nets through a few thousand evolutionary generations.
Then, the net was thrown into IC code, and tested. Did extremely well in
my rather biased opinion :)
-kr
===
Kate R.
Junior, ME '01, Math '02
Iowa State University
===
> Hi everyone,
>
> We've been reading up on the many different algorithms for mapping an
> environment using autonomous robots. Has anyone else been trying to figure
> out the best approach to map an area and it's obstacles using less than 32k
> of memory?
>
|
|
Message is in Reply To:
3 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
