For GameAI - algorithm or wing/hack it?
by Sean Brady · in General Discussion · 01/05/2012 (12:43 pm) · 5 replies
I am currently working on a GameAI portfolio, refining an obstacle tracing example at present. Anyway, I have the obstacle tracing working and refined enough to be usable in game but I am concerned about whether or not someone should stick rigidly to available algorithms or hack a new solution that works for their purposes based off the artificial intelligence concept rather than pseudo code of the algorithm. I have it working through hacking at it then clearing it up.
Ultimately, better to stick to known solutions or understand them and make a new one that suits yourself, the programming language and api being in development?
Would companies hiring look down that? If someone could shed some light on this issue for their own experience it would be greatly appreciated...
Cheers.
Ultimately, better to stick to known solutions or understand them and make a new one that suits yourself, the programming language and api being in development?
Would companies hiring look down that? If someone could shed some light on this issue for their own experience it would be greatly appreciated...
Cheers.
About the author
Professional mouth!, getting projects complete is the only problem.
#2
01/05/2012 (2:40 pm)
Cheers boss wil do... ;) 
#3
My recommendation would be use algorithms wherever possible, but at the inevitable point where the algorithm ceases to perform in the situations you put it in (pathfinding with dynamic obstacles, object tracing with concave hulls, etc.) then you've got to get creative and do whatever works for your particular needs.
So basically what Jeff said ;P.
01/06/2012 (6:13 am)
I'm not going to pretend to be particularly knowledgeable about how the industry works, but IMO, if you have an algorithm that solves your problem, you've every reason to use it. The thing is that in AI, you're trying to solve very tricky and ill-defined problems, where the results don't just have to be correct, they have to be human.My recommendation would be use algorithms wherever possible, but at the inevitable point where the algorithm ceases to perform in the situations you put it in (pathfinding with dynamic obstacles, object tracing with concave hulls, etc.) then you've got to get creative and do whatever works for your particular needs.
So basically what Jeff said ;P.
#4
If you do use your own algorithm (e.g. a 'Brady Search' instead of A*), then you must show why it is an improvement over the common algorithm.
If I was doing an AI portfolio, I'd look for examples of bad AI in current AAA games and come up with solutions. For example, the NPCs in Skyrim have a bad habit of blocking you ( http://www.cad-comic.com/cad/20111116/ ). If you had a good example on how to solve this problem, several companies would hire you on the spot. :)
01/06/2012 (11:49 am)
To be a good AI programmer, you need to know the common algorithms (A*, BT, etc.) but you also need to know when/how to use them.If you do use your own algorithm (e.g. a 'Brady Search' instead of A*), then you must show why it is an improvement over the common algorithm.
If I was doing an AI portfolio, I'd look for examples of bad AI in current AAA games and come up with solutions. For example, the NPCs in Skyrim have a bad habit of blocking you ( http://www.cad-comic.com/cad/20111116/ ). If you had a good example on how to solve this problem, several companies would hire you on the spot. :)
#5
Time to get cracking on it. I will put it up for critiques when complete (or when significant segment is complete).
01/07/2012 (6:30 pm)
Thanks for valuable information and heads up...Time to get cracking on it. I will put it up for critiques when complete (or when significant segment is complete).
Torque Owner Jeff Yaskus
jy games