will, i get the idea that it may be for the better if i just let the robot dance the roomba, 
the only other sollution i see, is using the robot as a input/output device en do all the math on a computer wich is connected with a wifi link.
although, i have been experimenting with the wavefront in visual basic, but (probably it's just me but,) it seems to take a much longer time to process the entire map on a pc then on the arduino, is this possible?
Your code on the device could simply be better optimized as compared to your v.basic code.
From my own research, the A* (A-star) algorithm may be what you are looking for. It takes a known map and a goal, and finds the shortest path that will get you there, while avoiding obstacles. I want to be able to do very similar things with my robot, but indoors.
Regarding memory, you may want to look into incorporating an EEPROM chip, or even an external flash memory card. I thought I was going to have to use a huge grid, before I did the math using a 1GB SD card.
Here's my take on the whole project (you can take it for what it's worth)
-GPS is going to help, as this will give you a totally external sensor to compare everything to. Look at Kalman filters to see what I mean.
-Wheel encoders are the way I am going, but I am indoors, on smooth surfaces. A small bump will cause these to drift significantly enough to throw your robot off course, and into the house...
-An IMU ($$$) could help, if you want to do the math related to that. (I may be moving in this direction myself)
-Regardless of what you use to determine location, you may want to drop the live mapping, and do it before giving the robot a task. Using SLAM is going to be tough, especially if you are limiting the whole project to just 1 arduino.