best solution for AI in T3D?

#1 Kory,

I am of the opinion that AI is VERY subjective. I don't think there is a 'single' solution that works for everything. Now, you did present some specifics and based on that... both UAISK and TAIK would meet your needs but, be prepared because neither solution will meet your needs 100%. You are going to have to tweek, re-tweek and tweek again no matter what solution you pick.

Here is the question you SHOULD be asking. Do you have the game play figured out for all of your 'levels' or areas? If you do, then you can go through any AI solution out there and figure out what best matches your game play mechanics. This is where level design comes in. Also, have you thought about the number of AI you need? Remember, each AI takes a specific number of cycles.... these cycles then 'slow' the system down, because you have to calculate AI, render the scene, manage a TON of other information... all within those same cycles.) SOOOO, if you need 100 AI bad guys that stalk, run, guard, attack, etc.... chances are you will NOT want a purely scripted solution. (Since, traditionally, scripted AI tends to use up more cycles as it 'thinks'.) Other things to consider, cover? concealment? Do they all use the same attack method? Do your AI need to be able to raycast and check for status of objects? (aka different weapons, health packs, etc.) Let's not even get into pathfinding cuz that is a whole other can of worms. All of these things need to be thought of prior to 'picking' an AI system. When I make a game level (and yes I make a lot of 'game levels' that don't see the light of day), I tend to lay out what I need the AI to actually do in that area. Then I use a combination of logic trees, or just good old fashion target nodes to make it happen. In the end, AI is all about game play. Once you figure out what you want to happen, you can better decide what will work best.

Ron

#2 I have gameplay logic though out for the whole game. I haven't put it in game yet. So it's all still untested theory. The concept isn't much different from the default T3D. Just a modify 3rd person camera. Gameplay similar to splinter cell mixed with traditional first person games. The difference would be how the ai plays. The enemies would be more like L4D, more melee based attacks. Most of the enemies would act similar, just different levels of speed, and aggressiveness.
As for ai, we figured no more that 25. Then one will respawn after one is killed. Good thing is enemies will only after the players, so no need for it to go after items.
I suppose the big issue is drawcalls and things slowing down the framerates.

#3 Sounds to me like you can get by with EITHER UAISK or TAIK then. Zombies and things like that don't really need to think that much so you are in luck. Your biggest issue will probably be adding additional animations to the stock T3D core and 'objective blockers' (aka path finding).
Thankfully, adding animations has been made much simpler in T3D. It was a major pain in earlier versions. As for AI path finding, Recast or Walkabout would likely work just fine. Either way though, there are going to be places were you will need to deeply understand each system, and it's going to be a matter of tweek and re-tweek then test and repeat. As for draw calls and such... too easy, just stack your texture info as compact as you can. Heck, my conifer pack was probably the worst at it (since it was my first pack) and even it calls a HUGE number of textures without maxing out the draw calls, the Africa Pack is much better and has Twice the texture requirements. You will get it with lots of testing and lots and lots of work. Hope this helps. Oh yeah... remember this. It's a game... Not a simulation. Cheat out your melee attacks. You don't need true melee to make it look and feel like melee. I reference Dead Island yet again... bad game (if not co-op) but it handled many many things well. Also, I know you want the community to input so, I will finish with this note.... but, contact me via my email brother. This and PhysX has been my latest 'pet project' so I have some more 'in-depth' info if you want to get into the nitty gritty.

Ron

#4 TAIK is a great tool but the creator, "Bryce" is mia. I was talking to him about the update and he disappeared.