But we still need the world internally consistent, so some gameplay concessions are needed to make it work. We decided how the game should play, and how the world should work, and inferred the rules of magic from there. But things got much better later on after I got an awesome writer on board and we chose to follow a top down approach. Tags game design / world buildingĮarly on, I had a very hard time with world building, as there were countless decisions to make at once. One of the problems with ANN is that they make having a behavior specific to each opponent harder (like having opponents that don't care about collateral damage, other that would only engage if they have overwhelming odds.).ĭate sam. I will probably start with handcrafted utility function parameters, then determine where ANN would help the most. It will "only" change the latest partĪnother possible use is to replace simulations with a trained classifier, so that the AI can estimate how (dis)advantageous is a given situation.īut we are not there yet. So the first possible use of ANN could be to determine how to score the options given the utility parameters, or I could handcraft them. I will probably use utility functions to score all options. The micro tactical AI will decide who to target, and in which tile to land. ![]() The tactical AI will decide how to assign group and where to dispatch them. The strategic AI will decide the overall plan (hold the objective at any cost, wait for reinforcements, try to kill as many characters as possible). Anyway, it does not change that much in the end: I will probably just use ANN for a small subset of it after all.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |