I'm toying with the idea of trying a neural network to have the bot 'make decisions' with but i'm kinda drawing a blank on how that would actually apply to some things.Quote:
What sort of AI will you be using? Im still working on my monster AI, and at the moment it works quite well, although its difficult to test without implementing the majority of the monster system.
Right now i'm just using datetimes and ints to control how often actions CAN occur and random chance combined with bot behavior types and difficulty to determine how often they DO occur... then actual actions have random chance and all that stuff in them and try to determine the best course of action (such as running through a few tries to determine which way to walk to get closer to their target coords).
Haven't done any high level ai yet though so they are still quite basic. Using a main bot thread to run all my checks and then splitting off into behavior types which splits off into types of actions (move/atk/etc)
Duelist bot isn't quite like a person but if you bump up the difficulty it puts up a hell of a fight.
I'm rewriting almost everything and taking it over to my hellmouth source (which could use a ton of cleaning) so if i get things working there i'll prob just release it as an entire source cause it really tends to run into every area of code... trying to keep it more self contained this time though.
ps. Yes, duelist bot is working now quite well.
What's your oppinion on the best way to do ai korv? I hate using sooo many if statements just to do something simple but something tells me any more advanced system will cause more problems... i do like the idea of a connectionalist structure though.. give weight to diff actions and have it spit out a choice based on the result/history of the network