Help with hopefully a familiar AI searching problem

j (
Mon, 2 Mar 1998 16:58:29 +0100 (MET)

Hello, i am a mechanical engineer, with sort of an advanced challenge in AI
understanding/programing. Never having advanced classes in AI programming,
i'm hoping
some of you can relate my problem to some of the freely published routines.
I know that what i'm trying to do has certainly been done before, and am
hoping that someone in this group has run across this before, and can point
me in a direction for more rsearch. Brute force searching the internet has
revealed hundreds of archives, examples, and sample problems--sorting
through will take some time.
The problem: a 2 dimensional (planar) search area, a simple square for now.
A target T(x,y) exists somewhere on the grid. An origin O(x,y) also exists
elsewhere on the grid (middle for now). Any number of searchers (points),
starting at the origin, move from the origin, and try to find the target.
The searchers will have limited travel, field of view, probability of
finding target, etc. weighting or penalties will be assesed to each
searcher for difficulty of terrain, field of view, ability to see the
target, etc. Eventually, the search area will be milti-dimensional, with
multiple moving targets, multiple searchers with differing searching
capabilities, obstacles, etc. --aaaack
Numerous classical algorithms have been published on this topic. I dont
have the time or funding to fully research and implement each one, i am
hoping that this problem is similiar to one that someone has seen before. I
have a feeling its in the chapter dealing with A*. A first thought is to
use fuzzy logic to direct the searchers towards the object, with feedback
between the searchers, and thier placing within the search area. A balanced
search pattern is the first step of this project. Random movement should be
minimized, each move should be strategically calculated, to better search
the most area in the minimum moves. Coverage of the search area is of
importance, but unlike the traveling salesman problem, may not be warrented,
if the target is spotted. Each searcher will have a polygon or circle that
is his field of vision, target too. Tons of imformation exist for polygon
detection and interference, showing when the target is spotted. Eventually,
with a working system model (hopefully one is already out there) this study
will be statistical analysis of the searchers ability in numbers.

Anyone with any experiance or information on algorithms i can study, please
dont hesitate to post here or to Any extensive
help will of course be credited in any publications. My main research deals
with fuzzy logic in mechanical systems, and this is a bit of a stretch.
Thanks in advance, Jay :-)

btw, if i'm way off base here, please post and tell me :-)