Forums

Adaptive techniques for GPS accuracy?

Started by thomasflynn16 February 16, 2016
Hi everybody, I have to do a research mini project on the following
techniques.

1- Kalman

2- Complementary

Even just some tips/links on where to start and maybe how I might be able
to add a practical element to my project (e.g a basic matlab script).

Considering my knowledge of DSP is somewhat limited, what fundamental DSP
aspects do I need to know before I can delve into these 2 techniques?

Background:
This question is asked by a final year Electronics student who has a very
very very basic understanding of DSP and a mediocre lecturer.
I know the basics of an IIR and FIR filter, High & low-pass filters, and
band-pass filters.





---------------------------------------
Posted through http://www.DSPRelated.com
On Tue, 16 Feb 2016 09:55:25 -0600, thomasflynn16 wrote:

> Hi everybody, I have to do a research mini project on the following > techniques. > > 1- Kalman > > 2- Complementary > > Even just some tips/links on where to start and maybe how I might be > able to add a practical element to my project (e.g a basic matlab > script). > > Considering my knowledge of DSP is somewhat limited, what fundamental > DSP aspects do I need to know before I can delve into these 2 > techniques? > > Background: > This question is asked by a final year Electronics student who has a > very very very basic understanding of DSP and a mediocre lecturer. > I know the basics of an IIR and FIR filter, High & low-pass filters, and > band-pass filters.
Need more info. What's the problem statement? If you're trying to do sensor fusion on a vehicle, complimentary filtering only makes sense if you know the orientation well. If you have to deduce (or refine) orientation from a mix of gyro, accelerometer, and GPS data, then you need some nonlinear extension of the Kalman filter -- at the least you need an extended Kalman, for best performance you'd need something like an unscented Kalman or a full-on Baysian particle filter. To get at least an intuitive understanding of complimentary filtering, assuming that orientation is known, you should: * make a Bode plot of the position noise from GPS * make a Bode plot of the position noise from an accelerometer after it's output has been conditioned to give position * think hard about where each one is superior. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com