Hi all, Ive just started my third year of a games programming degree. Along with other modules I obviously have a major project to do! Im really interested in doing something AI related, but don't really know where to begin! Some ideas I have are: A small RPG - with this idea I was concerned that with all the modelling, animation, texturing etc I would have to do, it would make me come across to potential employers as a jack of all trades and not specifically a programmer! A small real time crowd simulation ie A small street with traffic lights, cars etc where all the characters would have paths to follow, goals to complete and would take into account red stop signs, green go signs and cars etc - with this idea im little concerned it would be to big a task to undertake! What would you guys suggest? Any tips or help is much appreciated!

Quote: Original post by phelaphant
Hi all,

Ive just started my third year of a games programming degree.

Along with other modules I obviously have a major project to do!

Im really interested in doing something AI related, but don't really know where to begin!

Some ideas I have are: A small RPG - with this idea I was concerned that with all the modelling, animation, texturing etc I would have to do, it would make me come across to potential employers as a jack of all trades and not specifically a programmer!

A small real time crowd simulation ie A small street with traffic lights, cars etc where all the characters would have paths to follow, goals to complete and would take into account red stop signs, green go signs and cars etc - with this idea im little concerned it would be to big a task to undertake!

What would you guys suggest? Any tips or help is much appreciated!

Do an interlocking puzzle solver.

A 2D representation of some mechanism with interlocking parts/constraints (with and editor to create and change/define constraints/interactions of the components.

The do a planner which solves the puzzle (finds optimal solutions)


The visual aspect makes it easier to visualize the solution being made....

I know how you can come across as a programming beast...Write a graphics and a physics engine of your own. that'll knock their socks off!

btw, these books are the bomb:

Game Programming Gems (The whole series)
/and/
Real-Time Rendering, Third Edition (Hardcover)

I'm actually writing a 2D physics engine for my senior project in computer science, but if you don't think you can pull that off and you want something impressive I think you could find it in AI. My other idea that I almost went with for my project was to create a function approximater with supervised topological evolution. I ultimately decided against it because a) I couldn't get people to understand the significance in short presentations, and b) I wasn't sure my algorithm would work very well, so I went with the easier more flashy project.

I think people do really enjoy(read: are impressed by) learning AI if you can make it work, but if you want anyone to really take you seriously you have to go beyond ANNs and genetic algorithms, and start using things like polynomial least squares, or fancy regression or gradient descent algorithms. Basically, in a good project you take someone else's idea and make it just a little better.

However, if that's too difficult. This idea doesn't seem too bad:

"A small real time crowd simulation ie A small street with traffic lights, cars etc where all the characters would have paths to follow, goals to complete and would take into account red stop signs, green go signs and cars etc - with this idea im little concerned it would be to big a task to undertake!"

My suggestion for that would be to give each entity a goal location (probably at random), and then use a combination of A* for the high level street to street path finding, and steering behaviors for object avoidance, path following, and interaction with traffic lights etc.

As for the RPG idea, I wouldn't recommend it. An RPG is a very large amount of generally simple programming. ie it's will take a very long time and a lot of work to do and may or may not be that impressive. It could potentially be good, but the work per impressiveness ratio is probably not in your favor.

Anywho, that's my two cents, and if you pick something that seems a little over your head now, there's atleast a few people around here who would be happy to help you learn it.