Hello everyone,,,
I'm thinking about using genetic algorithms for the AI in my game. Does anyone have any experience in this area? Just so we're all on the same page, this is what I'm talking about.
What I'd like to do is have an enemy that can adapt to, or at least change behavior based on, the player's actions over the course of a single game. So for example, an enemy on level 20 would be much "smarter" than the same enemy from level
I'm not sure if this is possible though since normally the "evolution" part of the process takes hundreds or thousands of iterations, with small incremental changes each time. I could do something like record a playthrough and run an offline simulation to train the enemies, but that's not really what I want. I want them to start from scratch every game and learn based on that particular player during that particular session.
I haven't seen any examples of this in the wild, so like I said maybe it's not possible. But maybe there's some sort of hybrid solution with behavior trees or something else to speed up the learning process.
Thoughts?
Najam UlPosted Aug 29, 2020, 8:30 AM