Newb in need of some tips on a uni project

Hello. I hope this is the right place to post, sorry it's long.
I'm a total newb when it comes to robots which is causing some problems.

I'm doing a disso on robots, specifically line followers and genetic algorithms/techniques to make one work, evaluate methods, technical stuff. Gonna be using LeGOS, Eclipse. The robots I have to work with are the RCX [2 LEDs, 1 bumper sensor] and the NXT. Bumper sensor should reset the program weights to change the robot's behaviour, should it impact anything, I suppose.

So at the moment one idea is to experiment with various NXT sensors as a means to get it to follow this line. Might compare the two robots if possible, and senses. Other ideas are about how many sensors to use, how many artificial "neurons" [code weights] to use, what works well. i.e. Supposing two neurons to be a "simple solution", would 1 work, is 3 better, etc. Thick line to follow, thin line, shapes, etc. :-\

I'm supposed to look into feedback capabilities, such as IR or speech recognition software also. At the moment these don't seem possible or feasible.

I'm on a strict time limit, so I suppose I'm posting here seeking some thoughts, improvements or ideas on the subject. Perhaps some experiences from enthusiasts. It'd be such a great help. I'm still struggling to make it sound interesting!

Hopefully all that makes sense, I'm new at this.

p.s. Asking for opinions is counted as research.