Hi there! This is my first post to gamedev.net so please don't be too cruel ;) I am to submit a computing project at the end of this academic year (june 2005), and the predominant idea I have right now for this project is to create a game agent to act as a partner to the player in a 3d environment. The agent would perhaps be a dog that acts as an autonomous sentinel that could be sent away and gather information about enemies. In a typical fps game, the sort of information could be the enemies' position and number for example. Now I've done quite a bit of reading on AI, artificial neural nets, genetic algorithms, fuzzy logic and so on, with all that I don't really know where to start. I don't know what architecture my bot should be based on. Any help appreciated.

I think you would have more luck looking at robotic control architectures rather than the list of techniques you have mentioned.

ANN's, GA's, etc are all algorithms for learning. What you want is more akin to behavioral algorithms. In this case you probably need to look at subsumption architecures, state machines, and the reactive paradigm (I'd go with one of the first two).

The other thing is what Madster said. You need to specify all possible inputs and outputs your agents posseses. After that, you can worry about do based on the inputs and outpus (Perhaps a mapping from inputs to output?).