Differences

This shows you the differences between two versions of the page.

--- code:7_state_ekf_with_bias [2010/01/10 00:13] – laurenceb
+++ code:7_state_ekf_with_bias [2010/01/10 00:23] (current) – laurenceb
@@ Line 1: / Line 1: @@
+===== Bias tracking in Octave =====
+ This has been tested with three axes of accel and gyro, adding a magnetometer will give a dramatic improvement as absolute yaw determination will be possible. Interestingly this filter is still able to filter out yaw gyro bias errors if the IMU is tilted - i.e. any roll or pitch.
+In the first plot you can see an artiffically introduced large yaw bias being corrected reasonably well after the first few manovers. Angles are in radians.
+{{code:bigbias.png|}}
+ This second test uses raw test data with a sparkfun 6DOF IMU - i.e. actual biases. The intialisation of the filter was just guesstimated, and it takes a while to settle. However the values after 40 seconds are within 15% of the bias values obtained by interpolating between the gyro outputs when the IMU was stationary between movements.
+{{code:biases.png|}}
+===== GNU Octave code =====
 <code matlab>
 function quat_IMU_bias(beta,alpha,q,b)