Actions
Forum Thread Details
  • Replies 5 replies
  • Subscribers 54 subscribers
  • Views 447 views
  • Users 0 members are here
Related

Increasing accuracy in IMU

Former Member

I was trying to get roll ,pitch, yaw by applying direction cosine matrix , used accelerometer and gyro for roll and pitch, and magnetometer and gyro for yaw. I could get the readings perfectly but after rotating the my cicuit board in 3D space and bringing it back to original position i found that the roll and pitch remain almost same as original but there was a considerable change in yaw. Can anyone provide a solution to stabilize yaw???

  • I have watched IMU mature over the last four decades and it is not a simple problem to solve.

    In your case however, if you are going to limit your vertical movement, you can set up a set of states based upon your directional vectors and have them correct the signs of the true orientation.

     

    Just sit down and consider where you get the errors and then look at what you need to fix them. Usually, you can just change the polarity sign at a few select points.  From there you can setup the axial states you need to keep track of.

    Its not a fancy solution, but it is easy to determine and use in real time.

     

    I have seen some IMU software implementations that run into the million lines of code range, but you can do a fairly simple solution as long as you are not trying to circle the world or anything like that at high precision.

     

    Just a thought,

    DAB

  • in reply to DAB

    Can you please elaborate this a little. ???

    The problem is that for initial calculations some errors can be neglected but this error increases with time and will lead to larger unavoidable errors after small interval of time. I tried improving DCM and got it solved to much extent. But still the yaw keeps on drifting by minor amount even when circuit board is completely stationary.

  • in reply to Former Member

    Ok, You need to first verify your data to see if you are getting the correct input.

    You should verify the the calculations using a very simple test where you can monitor the actual position of the device and compare it to the IMU calculated position.

    If you are using sines and cosines, you can get discontinuities when you traverse through the 90 degree points, especially if you are using tangent angles for any of your calculations.

     

    Yes, when you do direction and movement adjustments based upon acceleration, you can build up a lot of error if your sensor data is noisy.

    You may need to do some heavy filtering on your data if that's the case.  That is why most of the new systems use GPS so they can reset the position points at the waypoints.

     

    Unfortunately, I no longer have the math skills to tackle this problem.  All I can do is suggest some testing you can do to see if you can isolate the source of your error.

     

    As I stated before, IMU implementations are not trivial.

     

    Good luck,

    DAB

  • Maybe you have already done it but, have you calibrated the sensors?

  • in reply to Former Member

    Do you mean removing the intial offset values by the caliberating ?? Or setting the Kp, Ki constants for DCM matrix ?? I have done both of this stuff ..