3.2 Attitude data from INS 
Attitude data, i.e. the roll-, pitch- and yaw- angles of the 
sensor body axis to the terrestrial fixed coordinate system can 
be directly measured by an inertial navigation system (INS) 
belonging to the equipment of every larger aircraft. The eva- 
luation of INS measurements is in most cases very difficult and 
depends essentially on the physical model which describes the 
systematic influences of a large number of physical effects. 
The analysis of INS data presented in this chapter examine only 
the registered attitude angles. Any systematic drift effects 
caused by remaining errors in the gyros of the INS are not 
taken into consideration. For this reason the filtered atti- 
tudes cannot be estimated as absolute values, but must be rec- 
tified for the systematic drifts. Then the estimated accuracies 
of the measured angles are understood as the precision of one 
reading of the gyro. 
The data for the analysis come from two different dynamic 
applications of INS. In the first example the attitude data of 
an INS from a NASA Space Shuttle mission are presented. The 
analysed data are registered during the metric camera experi- 
ment of the ESA/NASA's D1 Spacelab mission on 2nd December 
1983. The registrations took place every 2 seconds, the angles 
are measured in degrees. The attitudes refer to the earth-fixed 
Greenwich True of Date system. The registrations were subdi- 
vided according to 21 photo-strips consisting between 120 and 
800 points. As the estimated accuracies are very similar bet- 
ween the strips, only the root mean square values from all the 
strips are listed in table 4. 
  
Table 4: INS attitude data from Space Shuttle 
Estimated standard deviations in [deg] 
YAW PITCH ROLL 
observation noise On 0.0060 0.0048 0.0060 
ARI-model errors oce 0.0019 0.0012 0.0014 
filtered data ox 0.0027 0.0027 0.0031 
Correlation coefficients of filtered data  d-2 sec 
r(1d) 0.71 0.62 0.67 
r (2d) 0.46 0.39 0.47 
r{ 3d) 0.18 0.10 0.15 
r (4d) 0.13 013 0.13 
  
  
  
The mean ARI process order in this case is (6,1). The estimated 
accuracies are not very different between the three angles, so 
that a common measurement precision can be assumed. The estima- 
ted autocorrelation coefficients show a decrease within the 
first five seconds. 
87