Behaviour Trees seem fairly complicated and I haven’t had the time to learn them yet. Is it sensible to do all AI with behaviour trees, or does regular Blueprint (through a Character BP) work just as well? Is there a big advantage that behaviour trees have over Blueprint? Would I be better off just learning how to use behaviour trees instead?
Well… I worked quite a lot with AI’s. I toyed around with behaviour trees and as far as I know they are pretty neat in terms of practical reuse and less performance consumption. However, I don’t particularly like the tedious workaround feel, when working with these. Some steps seem just overly convoluted. I always wanted a lot of control over my AI’s, so I am one of the old AI-move-to blueprint users. I always add a lot of complexity to my stuff, without increasing the performance consumption. I have my little priority tree, which is kind of my implementation of Unreal’s behaviour tree, which consists mainly of Custom events.
The cool thing atleast with my approach is that I can create a very complex AI behaviour with sub trees consisting of, you guessed it, custom events. So by creating a lot of cross connections, I can have a nice AI doing complex stuff.
Just take, what I said with a grain of salt. Maybe I am just an idiot and it is totally wrong, what I am doing (and surprisingly working), haha :D.
Unreal has several different systems as far as I can gather - Behavior Trees, Pawn Sensing, AI Perception. Im a newbie myself but I opted for doing a combination of BT and Blueprint to get my AI working. I use the BT to drive the different behaviors and I get all my variables from the Pawn BP itself. The biggest benefit (aside from performance) is that you can reuse any task or service created on multiple BTs.
The A.I systems are not complete and are on hold for undetermined time;
If you can make your own, why not.
I would just make sure that it actually isn’t running slower than using what is already in place…