Game Development Community

Pathfinding libraries

by Geom · in Torque Game Engine · 12/29/2005 (9:53 am) · 11 replies

Hey guys,

I'm looking around at pathfinding libraries, to use in my game. Does anybody have any good/bad experiences with using any of these...


PathFind
http://www.cs.ualberta.ca/~games/pathfind/libpathfind/0.1.0/doc/

PathLib
http://pathlib.hildebrand.cz/pathlib.html
Notes: C. Docs are in Czechoslovakian (a drawback for me...)

PathEngine
http://www.pathengine.com/
Notes: costs about $1500 for an indie license

2DPathfinder
http://freshmeat.net/projects/lib2dp/, http://lib2dp.sourceforge.net/
Notes: C.

MicroPather
http://www.grinninglizard.com/MicroPather/
Notes: C++. Allows incremental world updates. Only two source files.


I am leaning towards MicroPather. It looks simple and I like the incremental update feature.

Thanks!
Geom

#1
12/29/2005 (10:02 am)
I just wrote my own A* system and navigation mesh object(in C++). But pathfinding is only one part in AI(if that's what you're getting at), and the AI pack seems to be on the doorstep of being released.
#2
12/29/2005 (10:58 am)
Hi Josh,

any particular reason you rolled your own code?

This is for an RTS game, so I'm not sure how much of the AI pack (besides pathfinding) would be useful to me. I'd really like to consider it though. After all it will be designed to work with Torque, which gives it a leg up over the competition.

GG guys, is there an ETA on the AI pack yet?

thanks,
Geom
#3
12/29/2005 (11:36 am)
Well there's a few reasons:
-As a learning experience.
-I enjoy AI related coding.
-The AI pack isn't out yet.
-There's no decent AI/path finding resources.
-My requirements for pathfinding(ie: area that needs to be covered) is huge, therefore I need a custom solution.
#4
12/29/2005 (11:50 am)
A* isn't particularly complicated - that's code that can be whipped out pretty quickly --- arguably faster than trying to adapt a third party library (though I admit I haven't used a third-party library for pathfinding, so I may not be correct in that).

It's the node generation, optimization, and smoothing out the path so it looks organic that is the trick.
#5
12/29/2005 (3:03 pm)
Just a quick FYI, the "AI Pack" (there are a couple last time I checked) is not under development by GG, so we don't have any official information about it/them.
#6
12/29/2005 (3:15 pm)
What makes Josh Moore think that "the AI pack seems to be on the doorstep of being released"?
I haven't heard anything about the pack for ages, anyone in the know?
#7
12/29/2005 (3:23 pm)
I agree totally with A* being fairly easy to implement and that the tricky parts are "It's the node generation, optimization, and smoothing out the path so it looks organic that is the trick."
However, the biggest problem for good, useable pathfinding is how to take other moving bots/players/dynamic obects into account.
#8
12/29/2005 (4:14 pm)
Phil Carlisle's .plan makes me think that. It sounds like all the tech is in place, but they just need to make some demos to show people how to use it. But don't take my word for it, that's just what I'm guessing.

And yes, A* is pretty easy to get going. However, having fast pathfinding with dynamic paths(ie: following a moving target) and intepolation/smoothing is, IMO, hard to get right.
#9
12/30/2005 (10:22 am)
@Stephen
Thx, my bad.

whoa -- I'm confused as to what AI pack we are talking about. I was talking about Justin Mette's work ( http://www.garagegames.com/blogs/1828/5854 ). Is Phil Carlisle working on a different pack or is it the same as Justin's?
#10
12/31/2005 (3:10 pm)
MicroPather looks really cool. I'll have to keep that one around for future projects...
#11
01/03/2006 (6:36 am)
Geom, it's the same pack. Check out Phil's plan.

Nick