The attitude data of the second example are registered during 
an aircraft flight with a Litton LTN-72 Inertial Navigation 
System of a Falcon jet. The registration rate was 10 Hz. The 
characteristic of the aircraft flight is very different to the 
space shuttle flight, which is much more smooth. This is ex- 
pressed by the ARI process parameters; here the ARI process 
order was only (2,2). The analysis refers to the central part 
of the flight, not disturbed by take-off and landing maneuvers. 
The results are presented here with reservation. Some discre- 
pancies in the results caused by unsteadinesses in the aircraft 
trajectory are subject of further research. During undisturbed 
parts of the flight the standard deviation of the ARI-predic- 
tion errors ce decrease below 0.0001 deg. 
  
Table 5: INS attitude data from aircraft 
Estimated standard deviations in [deg] 
YAW PITCH ROLL 
observation noise On 0.0023 0.0035 0.0030 
ARI-model errors Ge 0.0005 0.0013 0.0026 
filtered data ox 0.0011 0.0013 0.0017 
Correlation coefficients of filtered data d=0.1 sec 
r{id) 0.86 0.79 0.80 
r(2d) 0.57 0.40 0.41 
r(3d) 0.29 0.09 0.08 
r (4d) 0.09 -0.07 -0209 
  
  
  
The estimated observation noise on is in full accordance with 
other investigations from a stationary INS of same type. In the 
stationary mode the precision of the attitude measurement is 
about four times higher than in the dynamic mode, which was 
expected in advance. 
4. Conclusions 
This paper introduced autoregressive integrated stochastic pro- 
cesses for modelling the dynamic characteristics of the exte- 
rior orientation parameters of a sensor platform. The ARI-Model 
in combination with a variance component estimation enables the 
entire functional and stochastical description of the orienta- 
tion parameters. The main advantages of this model are the easy 
handling, the low number of necessary ARI process parameters 
and the dispensation of any a priori information concerning the 
stochastical model. 
The successful application of the ARI model to very different 
kinds of sensor orientation parameters improves the power of 
the concept. Comparisons to other methods for accuracy estima- 
tion demonstrate that the obtained results are realistic. 
88