1

Simply , how can I calibrate IMU unit ?

I read some papers about this topic and was wondering if there are any standard methods.

Paul
  • 1,278
  • 1
  • 14
  • 27
xsari3x
  • 137
  • 1
  • 6
  • 1
    Can you describe what the IMU is currently doing incorrectly? – Ian Sep 02 '14 at 20:28
  • @lan Ok , I'm trying to estimate roll & pitch , I used the following method : first calculate rp from gyro , rp from accele , then use 2d KF to fuse them , this I will have a clean r,p Next is to use 1d kf on accele x,y,z outputs then i calculated the standard deviation of accele outputs, if the next value of acc is bigger than the previous acc output by at least the standard deviation , this means that the system moves and then i start accepting the previously calculated pr , so I didn't calibrate it yet , thought that calibration will enhance results – xsari3x Sep 02 '14 at 21:03
  • @Ian check the comment – xsari3x Sep 02 '14 at 21:03
  • 1
    In general, these devices are dead nuts accurate, perhaps with small offsets. They shoud not require calibration other than maybe removing an offset. – Scott Seidman Sep 03 '14 at 10:43

1 Answers1

2

Inertial sensor assemblies suffer from a number of error sources, including offsets, scale factor errors and limited frequency response. With multi-axis sensors, you also need to account for alignment errors among the axes and cross-axis coupling. In addition, most of these errors are temperature-dependent.

How you deal with these errors depends very much on your application. Designers of advanced systems evaluate each of them and determine which ones can be dealt with manual calibration (e.g., axis alignment), and which ones are dynamic enough that they need be be dealt with in the Kalman filter (e.g., offsets and scale factors), adding additional terms to the system model to account for them.

Andrew
  • 560
  • 2
  • 10
  • 26
Dave Tweed
  • 123
  • 5