I’m currently studying the Introduction to AI class that Stanford University are providing for free. The most recent Units of study have been on Computer Vision and Robotics, two subjects that greatly interest me.
I race RC cars on a Friday night at Newbury RC Club in Greenham Business Park, and as a geek have wondered how easy it would be to get a self-driving car around the track
I had originally thought about trying to get the car controlled by Arduino (I have a Duemilanove), using Ultrasound sensors to detect obstacles (the track). However, having now studied the Computer Vision unit of the course, it seems that using a camera, and linear filters to detect obstacles will be a better solution.
There is some in car footage here:
The camera here is a little low – a higher viewpoint might offer better images. The track seems to offer easy differences for feature detection – plain grey carpet, with white drainpipe/yellow bollards for edges.
