Replies to This Discussion

DeveloperPermalink Reply by Ryan Beall on December 11, 2009 at 4:08pm

shake it and see what happes to the data.....put it in a steady state turn and see what happens.....There are a lot of physics that have to be accounted for. On the bench accelerometers work great as leveling devices but if they are moving or linearly acceleration (IE flying) they get confused as to where ground is.

Which is why you use gyros to help them out. Research Kalman filter or DCM. Not as clear cut as one might seem but still quite doable

• ▶ Reply

Permalink Reply by Eric Tamme on December 11, 2009 at 4:18pm

right.. ive read on kalman filtering. I fully understand the principle of normalizing data to compensate for jittery readings (the math im not so hot on though) and the predictive sort of filtering kalman does.

I havent built anything that does stabilization while in motion, just a simple balance bot using this sensor i got.

I did write a simple Processing visualizer that shows the sensor output on a graph and i shook it around, moved it for and aft without tilting it and stuff. I was quite surprised at how stable it actually was especially with some thing simple like summing 10 samples and shifting one place. (im not quite to implementing my own kalman filter yet)

Im working on a small indoor (converted freeflight) plane that i am going to sort of use as a test bed for working on my own stabilization system, just to learn... any how it sounds like there are many crashes in my future.

• ▶ Reply

3D RoboticsPermalink Reply by Chris Anderson on December 12, 2009 at 12:24am

This question has been asked so many times we even have a FAQ on it on the front page!

• ▶ Reply