23 ms·

Poor Man's Explanation of Kalman Filtering [pdf]

59 points by kqr2 16y ago

axiom 16y ago
Wow, this is really well done. Everyone should go read this now! Kalman filtering is one of these techniques that's used everywhere and yet most engineers/developers don't even know it exists (let alone the math behind it, which is fairly straightforward undergrad level stuff.)
gvb 16y ago
"Taygeta Scientific Incorporated, has acquired the rights to republish the report." http://www.taygeta.com/kalman_book.html http://www.taygeta.com/kalman_book.html Has anyone bought a copy? I assume it is typeset rather than being a very difficult to read fifth generation photocopy scanned to PDF.
- kjhgfghjk 16y ago
  if only there was some way of filtering the noise
- smanek 16y ago
  Offtopic: Why was kjhgfghjk's joke (this comment's sibling) killed? It's funny, since the scan of the article is very 'noisy' and a Kalman filter is a possible solution to help clean it up (probably not the best, but still ...)
  - gjm11 16y ago
    I would say that in more than half the cases where I see a dead comment, there is nothing obviously very wrong with the comment. Sometimes it's dead because the user who posted it is auto-dead-ified -- everything s/he posts is killed -- but in most of the (few) such cases where I've looked back over the user's comment history, I haven't found anything very awful that they got put in that state for.
- gvb 16y ago
  Tip: highlight the nasty text: the PDF was done with OCR, so the text is magically cleaned up using "select".
dlnovell 16y ago
Awesome! I worked on the inertial navigation system for Insight Racing's entry into the 2nd DARPA Grand Challenge as my senior design project in college. We implemented a Kalman filter, but I'm not sure we got all the bugs out as it didn't work very well. I sure wish I had known about this little doc at the time!
- dlnovell 16y ago
  Actually, I should probably clarify: _our senior design team_ couldn't get the inertial navigation system working very well. I'm pretty sure they took what we worked on and fixed it as they finished somewhere in the middle of the pack. It was easily the coolest project I've ever worked on.
brisance 16y ago
Dan Simon's 2001 article is just as good, if not better IMO. http://web.cecs.pdx.edu/~greenwd/kalman_overview.pdf http://web.cecs.pdx.edu/~greenwd/kalman_overview.pdf
joeyo 16y ago
Welch and Bishop also have an excellent introduction: http://www.cs.unc.edu/~welch/media/pdf/kalman_intro.pdf http://www.cs.unc.edu/~welch/media/pdf/kalman_intro.pdf
mturmon 16y ago
If you just want the equations, try: http://www.ece.cmu.edu/~byronyu/papers/derive_ks.pdf http://www.ece.cmu.edu/~byronyu/papers/derive_ks.pdf And for the EKF: http://www-npl.stanford.edu/~byronyu/papers/derive_eks.pdf http://www-npl.stanford.edu/~byronyu/papers/derive_eks.pdf
andrewljohnson 16y ago
Here is an open source Kalman filter (written for my iPhone app) by a very smart friend of mine: http://github.com/lacker/ikalman http://github.com/lacker/ikalman It's in C.

59 points by kqr2 16y ago

axiom 16y ago

Wow, this is really well done. Everyone should go read this now! Kalman filtering is one of these techniques that's used everywhere and yet most engineers/developers don't even know it exists (let alone the math behind it, which is fairly straightforward undergrad level stuff.)

gvb 16y ago

"Taygeta Scientific Incorporated, has acquired the rights to republish the report." http://www.taygeta.com/kalman_book.html http://www.taygeta.com/kalman_book.html Has anyone bought a copy? I assume it is typeset rather than being a very difficult to read fifth generation photocopy scanned to PDF.

kjhgfghjk 16y ago
if only there was some way of filtering the noise
smanek 16y ago
Offtopic: Why was kjhgfghjk's joke (this comment's sibling) killed? It's funny, since the scan of the article is very 'noisy' and a Kalman filter is a possible solution to help clean it up (probably not the best, but still ...)
- gjm11 16y ago
  I would say that in more than half the cases where I see a dead comment, there is nothing obviously very wrong with the comment. Sometimes it's dead because the user who posted it is auto-dead-ified -- everything s/he posts is killed -- but in most of the (few) such cases where I've looked back over the user's comment history, I haven't found anything very awful that they got put in that state for.
gvb 16y ago
Tip: highlight the nasty text: the PDF was done with OCR, so the text is magically cleaned up using "select".

dlnovell 16y ago

Awesome! I worked on the inertial navigation system for Insight Racing's entry into the 2nd DARPA Grand Challenge as my senior design project in college. We implemented a Kalman filter, but I'm not sure we got all the bugs out as it didn't work very well. I sure wish I had known about this little doc at the time!

dlnovell 16y ago
Actually, I should probably clarify: _our senior design team_ couldn't get the inertial navigation system working very well. I'm pretty sure they took what we worked on and fixed it as they finished somewhere in the middle of the pack. It was easily the coolest project I've ever worked on.

brisance 16y ago

Dan Simon's 2001 article is just as good, if not better IMO. http://web.cecs.pdx.edu/~greenwd/kalman_overview.pdf http://web.cecs.pdx.edu/~greenwd/kalman_overview.pdf

joeyo 16y ago

Welch and Bishop also have an excellent introduction: http://www.cs.unc.edu/~welch/media/pdf/kalman_intro.pdf http://www.cs.unc.edu/~welch/media/pdf/kalman_intro.pdf

mturmon 16y ago

If you just want the equations, try: http://www.ece.cmu.edu/~byronyu/papers/derive_ks.pdf http://www.ece.cmu.edu/~byronyu/papers/derive_ks.pdf And for the EKF: http://www-npl.stanford.edu/~byronyu/papers/derive_eks.pdf http://www-npl.stanford.edu/~byronyu/papers/derive_eks.pdf

andrewljohnson 16y ago

Here is an open source Kalman filter (written for my iPhone app) by a very smart friend of mine: http://github.com/lacker/ikalman http://github.com/lacker/ikalman It's in C.