Code Snippets->Motion Planning
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 26th Sep 2003 1:00
Hi,
I wrote a tutorial about motion planning using potential fields, an extremely fast pathfinding routine, which was of my interest a certain time.
Naturally it is written in english with sources etc.
Feel free to comment.
http://www.thequest03.de/Tutorial%20-%20Pathfinding(english).doc
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 27th Sep 2003 12:53
Come on, guys...
At least give it a try.
H
Hamish McHaggis Joined: 13th Dec 2002
Location: Modgnik Detinu
Posted: 27th Sep 2003 16:30
It's a very good tutorial, I didn't have time to read it all but I'm sure anyone who is doing a RTS game will find it invaluble.
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 28th Sep 2003 2:32
At least someone says something .
I will upload a screen tomorrow.
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 28th Sep 2003 20:35
Here's a screenshot:
D
David T Joined: 27th Aug 2002
Location: England
Posted: 28th Sep 2003 20:46
I'll have a read

hmm, I have seen your signature quatoe before...
D
David T Joined: 27th Aug 2002
Location: England
Posted: 28th Sep 2003 20:51
From a short read this looks really interesting. Thanks )
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 29th Sep 2003 0:14
It should be interesting for you, as you are making an rts ...
M
M00NSHiNE Joined: 4th Aug 2003
Location: England, UK
Posted: 29th Sep 2003 11:39
I wonder if theres some way of combining both this and a* to produce the ultimate pathfinding algorithm?
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 29th Sep 2003 23:21
You could try
One possibility would be, to precalculate the whole field, with a type of waypoint system and then to calculate the way how to dynamically switch on and off attractors with a*.
For the local pathfinding, the algorithm posted by me would fit.
M
M00NSHiNE Joined: 4th Aug 2003
Location: England, UK
Posted: 29th Sep 2003 23:24
Well the fact that youve said its a possibility is a good start, now it just needs coding.....

Not really my area of expertise this pathfinding lark. Ho hum.
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 29th Sep 2003 23:29
Well, at the moment I'm trying to solve the local minima problem with a completely new and more complex approach based on the idea of potential fields(for optimal results).
But I think, I'll need a while .
M
M00NSHiNE Joined: 4th Aug 2003
Location: England, UK
Posted: 29th Sep 2003 23:33
Yep, suppose the local minima thing is the only thing stopping this technique being useful for RTS, which is what im planning to do. Good luck with it anyways
D
deckard Joined: 11th Apr 2003
Location:
Posted: 30th Sep 2003 11:41
Absolutely brillant idea ! I suppose, there is no need to calculate
sum (big greek E) including all objects, but only for some nearest
objects. You can also experiment with other functions such as
v#=k1#*2.0^(-distance#^2)
v#=k2#*(distance#+.1)^(-1.3)
if distance#<1 then v#=1-distance# else v#=0
and other simple functions with lim->0 when distance->infinity
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 30th Sep 2003 12:26
You're right; you do not have to calculate all the influences for one agent(because the influences are nearly 0 in a certain distance).
Another formula would be:

alpha*e^(-gamma/2*((x-a.x)^2/s+(y-a.y)^2/o))

You would have more control on the shape of the repulsor.
s and o could be used to define the length and height of the ellipse.
K
KamaKase Joined: 29th Oct 2002
Location:
Posted: 27th Oct 2003 20:16
What's an algorythm.....
Aaaaw, why is everything so complicated.
D
D Man Joined: 3rd Oct 2002
Location: Germany
Posted: 27th Oct 2003 21:17
Thats the price for the speed .
R
rob p Joined: 24th Sep 2003
Location: Philly, PA, USA
Posted: 28th Oct 2003 17:46
Excellen tutorial! This will definitly come in handy. Thank you