I'm trying to work on a project which is based on navigation of a robot which is equipped with a web camera and the robot will have a ARM based board like Raspberry Pi or BeagleBone or BeagleBoard (whichever is good). What sort of line detection algorithm can be used so that it can navigate pathways without collisions or bump into obstruction. Or is a map needed for navigation irrespective of whatever robot is being used. Please answer this as i'm new to robotics as well as GIS. I'm doing this project for my undergrad finals. Also i would like to learn a bit about UAVs used for indoor navigation as done by MIT using localization Youtube video here
Having a Raspberry Pi myself, I have been amazed on the variety of solutions that can be done with a £25 ($35) computer.
Update: This can be done and here is some proof:
there is a video of this with the wireless keyboard http://dexterindustries.com/blog/2012/08/01/raspberry-pi-and-the-lego-mindstorms-nxt/
On Friday 10th August 2012 there was a video posted on the official news/blog for Raspberry Pi The video (English or German) shows from a very basic setup 17 LED's using the GPIO to very advanced (controlling a motor vehicle's engine rpm)
http://www.raspberrypi.org/archives/1743 with the videos
NOTE there is a section where they are controlling small-robots. So it can already be done. The challenge is to make the robots aware of their surroundings in real-time - has the raspberry pi got enough computer power (Memory!) for this??
I am fairly sure QGIS would run on Ubuntu (slowly when processing) on an Raspberry Pi https://wiki.ubuntu.com/ARM but not tested.
RAM can be added to Raspberry Pi - 48GB in fact http://twitpic.com/ak51h8
There is a Raspberry Pi (Beta) Stack Exchange http://raspberrypi.stackexchange.com/
or ask on the official forum site http://www.raspberrypi.org/phpBB3/index.php
GeoFencing might be a good solution on a very small (indoor) scale http://en.wikipedia.org/wiki/Geo-fence
Concerning the project of Indoor UAV , these UAV are using 20 rangefinders to measure the distance to obstacles in real time to avoid collisions with walls ... , but you will need much less than 20 rangefinder, if the robot can run with embed indoor map , you can use magnetometer to determine the direction of the robot, i have not an idea of the constraints of measuring the traveled distance by a robot, but if you have ( traveled distance + direction ) at real time you will have real time location.