I was also playing with the idea of building my own robot vacuum (I think roomba's are more like robot sweepers...), but I couldn't resist buying a roomba 5105 (one of the first models right?) for just 25 euro's!
I was planning on using a hand vacuum (they run on 2.4-3.6V) an arduino as brain and 3 sonars+bumper as sensors and a 1s lipo as battery (either all electronics at 3.3V or using a booster for the electronics).
But I think I would be smarter to use the 5105 as a platform to test obstacle avoidance and I also want to incorporate mapping (in my room it is not necessary, but I want to be able to beat my moms philips homerun!).
I was wondering if it would give this project a higher success chance if there were a couple of different approaches?
The big hurdle seems the choice of uC and what platform, right?
I'd say just about any roomba could be used as a platform, we could always change or add the necessary sensors right?
The final goal is to design and build a superior vacuum bot for the same money (or less off course), that should be the second step, first we need to find out how to make the bot aware enough of its surroundings and a brain capable of processing the "awareness" of the bot.
Just my thought on the matter
I will be dissecting the roomba as soon as I get a working battery...