Graceful Degradation and Recovery from GNSS Interruptions Figure 6. Time history of the attitude error (green square) and attitude blender's covariance (red triangle). The covariance is calculated using (9). Figure 7. Time history of attitude blender weights B1a and B2. Blender weights on the GNSS/INS solution (green square) and weights on the AHRS solution (red triangle). seen in the time history of the attitude blender weights plotted in Figure 7. This figure shows the diagonal elements of the blending matrices B1a (green square) and B2 (red triangle). Recall from Figure 4 that B1a is the attitude blender weight on the GNSS/INS solution and B2 weighs the AHRS solution. When the GNSS signal is lost, 10 the weights on the GNSS/INS solution go to zero while the weights on the AHRS solution become unity. This transition from zero to unity (or the other way around) occurs gradually. Thus, it can be seen that the blender can automatically exclude a failed sensor (or sensor system) and reintegrate it once the failure has been removed. IEEE A&E SYSTEMS MAGAZINE SEPTEMBER 2017