Create a method to generate different paths deterministically from start
position to end position. Compute which one is best, and keep it. Generate
more paths. Compare against each other and the previous best one, and keep
best one again. Keep doing this until you 'time-out'. This will find a
good solution, but not guarantee a best solution (and note that it is very
close to, but not quite, a genetic algorithm. It is more a typical example
of old style heuristics).
If you need a best solution, plan on it taking 100's of years in many
non-trivial cases, the problem can be very intractable due to the enormous
number of non-linear paths (e.g., path A + Path B != Path (A+B))....hehe
THAT's why genetic algoritms are the typical way to go, which makes me
wonder why any project supervisor for a COMMERCIAL company would ever
restrict HOW you got a problem solved for the hell of it.
Are you sure you're not a student? : )
[==P==]
Pohihihi said:
It surely would have a school project 5 years back when I was in school
but for now it is a project for me. While posting first time I was
assuming that someone will surely take it as a school assignment. Can't
complain cause I would have taken it that way myself.
Anyways, so the usual implementation we see for this algorithm is via
branch&bound or some kind of genetic algorithm. Google is full of all
those implementation. Basically we are in a project that is related to
hydroplant optimization. Don't ask me how that work cause I am having a
part in the whole project. Our algorithm expert has some argument about
not using regular/populer (and easy) method. Now for me I have not
implemented TSP without B&B so I am very much like a student right now in
knowing how to do it.
All this said, now question is, can you help me?
Thanks,
Po
