Which engine is better for AI behavior customization?

[laminutederire]

Hi there everyone!

For a project at school, I might need to create a mini game with a large enough size map (like an historical center). It would be an open world type of thing.

However, since I don't know any AI engine, I wondered which engine should be the best to customize the AI behavior? I would need to have a very specific behavior for it, to make it good enough as an experiment.

(It would be a non-profit type of thing where graphics will be subsidiary at best so I should have the choice between "classic engines" such as UE4 etc. I think large enough studios like Ubisoft might help us a bit if we ask kindly enough and present the project correctly)

(Second footnote, I know I should build my own engine for this, but it would be too time consuming to be done before the deadline)

Thanks in advance to any answers!

And while I'm at it, happy new year to everyone

[AlexM]

Are you supposed to write an AI solution yourself? If so would using an engine like Unreal be considered allowable? 

Unreal has some pretty impressive AI authoring tools but without knowing the specifics of the problem you are trying to solve I can't tell you if it's appropriate or not. 

[laminutederire]

Are you supposed to write an AI solution yourself? If so would using an engine like Unreal be considered allowable? 

Unreal has some pretty impressive AI authoring tools but without knowing the specifics of the problem you are trying to solve I can't tell you if it's appropriate or not. 

it would be to develop a game with an AI respecting certain coupling algorithms. The idea behind that is to gimmick some human behaviors and do tests in neurosciences with those programmed coupling in interaction with the player.

i don't know if that's clear?

for example you could walk with a kinect and the AI will have its step synchronizing with yours. (Bad example since it's hardly doable to walk long enough with such an équipement,  but it gives the type of coupling application.

Thing is that it all comes down to the AI respecting some differential equations and it being able to change the parameters of the equation to adapt its behavior.

[AlexM]

I think the important thing to take away is if you use some type of framework then you are working within the contraints of said framework. There's a point where if you design is unique enough then modifying an existing framework is more work then writing one from scratch. 

As for your particular problem, I think I would say I don't understand it enough to recommend a course of action. I'm not familiar with coupling algorithms. That said, Unreal engine is fully open source which is good for interfacing unsupported hardware like an IR camera. I just don't know if it's faster to work from scratch or not. My gut says using the engine would be easier overall. 

 

[laminutederire]

To my mind the engine would be easier overall because it would allow using a lot of assets and easier for the rendering part.

I think I'll go for UE4, even if the game will probably end up portal-esque

[clankill3r]

I have a book called "programming game AI by example" and one by o'reilly called "AI for game developers". The first book is really good, the book by o'reilly is not so good. I really recommend getting the first one.