I was hoping to get some ideas about a senior project. I'm finishing up my bachelors degree in computer science, and really hoping to go into gaming (or graduate school...kind of conflicting goals I guess). I've already finished two internships in it, and really have loved development side. But it was scripting stuff, not deep programming. So I am getting ready for my senior project, which is pretty huge. It is supposed to take 4 months of 12 hours of work a week (I have given myself an extra 2 on top of that to make sure I can graduate on time). It requires me to propose and complete a project outside of the curriculem, which is very focused on programming fundamentals. This schools theory is teach them to be good programmers and they will be able to learn whatever they want when they graduate. So lots of math, algorithm theory, software engineering, programming theory, computer hardware theory, etc. Nothing high level like AI. I would like to work on something that will be useful to me if I work in video games, like artificial intelligence. But I such a huge topic. I'm currently researching, but it is a massive topic to choose any single aspect and know if it will both useful and enjoyable to work on twelve hours a week for the next six months. So I was hoping to get some suggestions. I would like to work on something then implement in a simple graphical prototype. What aspects AI have you work with and enjoyed? What has been useful? What are some ideas about projects? I would really appreciate any feedback. Thanks.

Try to implement an actual game worth playing, not just a prototype. Use XNA or an existing game engine, that focuses on being easy to make look good.

For an undergrad project you usually implement existing algorithms. A possibility that would be quite useful starting out in AI is implementing the steering algorithms here:

http://www.red3d.com/cwr/steer/
http://www.red3d.com/cwr/boids/

Make a geometry wars clone and implement some of those algorithms for the AI.

Combine use of A* pathfinding with Influence Map analysis to generate the 'map'.

Basic mechanics for how the threats work (ie- guns with terrain cover scheme)


2D graphics to present the simulation and make visible the analysis conclusions.

User input to place/move the enemies and the targets.

That's not a bad idea. For information, look up Damian Isla's work with "occupancy maps". I believe he wrote a good article about it in one of the AI Wisdom books.

That is interesting. So if I understand right influence map analysis is an technique that basically gives weights to different parts of a board or map. A* is for finding shortest path.

But by combining those two you get a path that isn't just shortest distance, but also optimal from some other perspective (tactical in this case). For instance paths that keep you behind cover are considered 'shorter' by the A*, but so are paths that accomplish an objective.

Is that about the gist of it?

A* figures out the shortest possible path (A) according to movement costs (F) and a qualified guess (H):
http://www.gamedev.net/reference/articles/article2003.asp
That article use the Manhattan distance as heuristic, but people here are suggesting occupancy maps as a means to compute the cost (in H).

Quote: Original post by riplikash
Is that about the gist of it?

Yep. In a nutshell. There's plenty you can do with combinations of those.

..or on the other end of the AI spectrum, some kind of neural network / GA implementation.