Awhile ago, someone from RobotShop suggested I use a Kalman filter to determine my robot’s position over time. At this point I have a ZiChuan compass and a GPS receiver. I plan to add a gyroscope and an acceleromenter.
My problem is in understanding how to set up the various vectors and matrices. I’ve looked at a lot of articles on the Web, but most seem to assume areas of math, in particular Linear Algebra, that I’ve now forgotten or never knew. The articles I’ve found don’t demonstrate how to set up the inputs in sufficient detail for me to understand. OpenCV comes closest to setting up a Kalman filter and has functions to do the processing.
So, what I’m looking for are suggestions on how to approach filling in the areas I don’t understand. What parts of Linear Algebra are essential to using the Kalman filter?
Thanks for the reply. I took a look at all the stuff I’ve downloaded and found several articles that give me an idea how the equations work, but I’m nowhere near enough to using the Kalman Filter yet. Most stuff seems to be about how the filter works. The articles below do set up the filter, but without the detail I’d like. Each time I read them I understand a bit more, though. Much of the math is Linear Algebra.
Some references:
Two different articles by Dan Simon both tititled Kalman Filtering.
ive tried making my own kalman filter code and want to test it. to do this i need some method that will give me the “true” movement curve which i will then compare with what i will get using the kalman filter. any tips on how i can get this “true” movement curve?
Kalman filters are certainly becoming more and more popular for robots that needs to determine absolute orientation and positioning. If you want to understand the derivation and the complete theory, you would need to do quite a bit of research, and the level of math is certainly more at the collegiate / university level. Another approach would be to find code which has already been written and successfully tested and modify it to suit your needs. We currently do not have sample code to provide, though we welcome users to post their pseudi code here and elaborate on the theory. We will also post additional information we come accross or use.
It depends largely on what you are looking to have your kalman filter acoomplish - is it helping to balance a robot using inputs from accelerometers and gyros? Usually you can find a reference using an additional (absolute) sensor such as a tilt sensor, or even a weight attached to a potentiometer. Can you describe your setup?
