Hi out there!

I am looking for a navmesh creation tool. I saw some products from Kynogon navmesh greator, xaitment xaitMap NavMesh tool and engine and Havok AI navigation mesh tool.

What navmesh tool would you recommend to use? I have a 3D multilevel building with lots of stairs in it. xaitMap claimed to generate a nav mesh also over stairs Does this hold also for the others?

xaitment's xaitMap AI SDK was the only one I could directly download from the webpage: http://www.xaitment.com/english/products/xaitmap.html

I could make a lot of tests and I was really impressed by the speed of xaitment's navmesh generation. The size of the map is unlimited and with the sector loading it works fine. The memory consumption while generating the navmesh was small. The other engine I could not test, because of their missing download.

Does anyone has some recommendations?

[Edited by - oliverkriese on October 26, 2010 5:33:30 PM]

http://code.google.com/p/recastnavigation/

Didn't you just post this?

(Also, you wouldn't happen to be astroturfing for xaitMap, would you?)

Quote: Original post by jyk
Didn't you just post this?

(Also, you wouldn't happen to be astroturfing for xaitMap, would you?)

Yes I did, but the focus is a different here. I am looking for professional tools that I can test or download a trial version. In the past we implemented our AI by our won and we failed again. I am tiered of doing this again and want to use a good middleware, that solves my basic problems and reduces the project risk.

But Havok and Kynogon I didn't get a free eval version yet and other good AI engines I don't know. And Open Source is not an option if you are responsible for a team and a big development budget. You have no support and this is what I have learned over the years, don't use software no one feels responsible if you are just before a critical milestone.

So my question in this forum is, are there any other AI engines available that I can download and test?

Nope. AI middleware, for the most part, is either non-existant or sucks all to hell. You gotta make your own.

That being said, go buy all the AI Game Programming Wisdom books. Those give you a great start in a lot of areas -- pathfinding included.

Quote: Original post by InnocuousFox
Nope. AI middleware, for the most part, is either non-existant or sucks all to hell. You gotta make your own.

That being said, go buy all the AI Game Programming Wisdom books. Those give you a great start in a lot of areas -- pathfinding included.

Hi, is it really so worse? ok - movement I can imagine because often it doesn't fit to the game design and it is hard to adjust / modify a middleware. But what about Navmesh generation? and pathfinding? these are standard algorithm like A* and if someone has optimized them, it should help.

Or do you have a different experience?

What about scripting or FSM editors? I thought it would be a nice thing to have a good IDE like visual studio that helps me to create my content faster and that I can keep a better overview of it. Am I wrong?

Xaitment has a graphical AI authoring tool. However, the problems with behavior and middleware are actually worse than navigation issues. A grid is a grid is a grid... knowledge representation in world models and behavior modeling vary so widely that gluing something like a middleware product onto it is difficult.

Quote: Original post by InnocuousFox
Xaitment has a graphical AI authoring tool. However, the problems with behavior and middleware are actually worse than navigation issues. A grid is a grid is a grid... knowledge representation in world models and behavior modeling vary so widely that gluing something like a middleware product onto it is difficult.

You are right, knowledge representation is not that easy and very unique from game to game. But the approach from xaitment is new to games and sounds interesting. I thought of testing it for a small application and see how it works. The guys from black and white used machine learning and making an ape going crazy was real fun :-)

But modeling a FSM with tools instead of coding it like "if-then-else" (or switch) is reasonable for me. While using UML for your game design or specifications I learned that diagrams and graphics explain more than 1000 words. This is what I like at xaitment's approach. It is just a tool - not content.

So from that point of view you are right -ready-to-use content doesn't make any sense - maybe in world like SAP :-) but not in games

Quote: Original post by oliverkriese
But modeling a FSM with tools instead of coding it like "if-then-else" (or switch) is reasonable for me.

This is a common misconception. Just because you have a pretty graph doesn't make it any easier or faster to develop. Edit and continue support helps though!

However, if you're planning to scale up, you should roll our own behavior tree.

Alex

I've asked the moderator to look into this issue. I believe Oliver was hired to promote xaitment (or is actually an employee). It's not just this thread but the other he started.

Luckily, it should be easy to find out. Oliver, here's your chance to come clean and avoid any further PR blunders from your unethical marketing.

Alex