Fortunately, though, I am currently taking the Udacity "CS73: Programming A Robotic Car" being taught by Sebastian Thrun (**broken link removed**), which aims to give you the skills to "Learn how to program all the major systems of a robotic car from the leader of Google and Stanford's autonomous driving teams".
Yeah - I missed the first iteration of it, but when I saw it was running again, then found it was a "self-paced" course, I decided to try it out; right now I am beginning the last programming question before the homework on unit 2. Last night it didn't make any sense to me, so I am hoping it makes better sense here in a few minutes (been procrastinating on it a bit - but hey, self-paced, right?). I think he is wanting you to implement a 1D Kalman using matrix math (as I am sure the homework is to do the same, in 2D - but I haven't checked), but it didn't seem clear to me last night.
Welp - just got past it; re-watched the video and realized that he wanted us to implement a multidimensional Kalman - not 1D. Once I understood that (and the matrix class he supplied), it was relatively straightforward to plug-n-chug the pieces in place to implement the Kalman properly. Now on to the homework...
Multidimensional Kalman filter is one of the most important tool to develop mobile robot: I remember it when I studied, this kind of filter was in a lot of exams
ok just a thought but, how about a remote controlled mower, the idea being you drive it around by remote once and it saves the info in memory so next time it just follows a save path all you would need is decent encoders ect on the motors. or is this a stupid idea
Yep, encoders are good only if you are on a good ground, but they are not accurate if you are in a sandy terrain because wheels can slip and give bad readings.
I think the dog fence is the only suitable solution at this moment in addition to a non-randomly path.
I was thinking about how hard drive heads know their location by address marks on the disk. One can use that same sort of idea with a lawn by encoding data using magnets. Either the space between them or the polarity.
As I said it was a rough idea. I expect it could be worked out several ways.