Thoughts on using machine learning for enemy AI combat system

Started by
3 comments, last by tberry11 3 years, 5 months ago

Hello,

We have a role-playing game as a graduation project that will have similar combat system to souls games series (Dark Souls) but with enemies that will have human-like behavior through machine learning to be somewhat more realistic and more challenging.

What are your thoughts about this technique or field in game development AI?

Have you tried any of these techniques before?

Do you know of any projects that have similar approaches for enemy game AI?

Thank you

Advertisement

it's been done before, yes, in RPGs too, so there's hope for u as well ?

your biggest challenge will be to code up an agent that can learn and make decision purely based on what has been learned (as opposed to programmed), and maybe from this make up entirely new decision that u don't anticipate to see in the game. That would be pretty cool.

So advice, test it first in a sandbox and then deploy, but please make your findings public so we can benefit as well ?

i think these guys did it, and they had a kickstarter… , might be worth checking them out: (if u can find info on them) https://en.wikipedia.org/wiki/Planetary_Annihilation

Have fun ?

From an AI-driven actors perspective two things matter evaluating what another AI-driven actor or you (the player) did and what another AI driven actor or you will do. If we talk about the first issue all games cheat (or the AI cheats if you will, like if you take a game like Starcraft the AI player knows where you are on the map with no scouting), if you want realistic behaviour you must give up on cheating. Like if you`re building an RPG you want the AI actor (him) to `know less. You want him to observe what you (the player) are doing (notice/track the position of your legs, arms, weapons, your position in relation to other relevant objects on the map).

My project`s facebook page is “DreamLand Page”

@Ranyo13 This is an interesting and highly complex topic at the same time. Many people will say its not possible

to combine machine learning or deep learning and game AI. This article below that I came across will shed some light on some of the challenges those who went down this path have faced in the past and how they ended up facing a dead end.

https://www.quora.com/How-can-a-video-game-enemy-AI-be-improved-with-deep-learning

This topic is closed to new replies.

Advertisement