interval. For a low performance system the noise level
will be above the one second value.
Error in System Accuracy Class
nav. grade low accuracy
Attitude pitch & roll; ^ azimuth pitch & roll; azimuth
1h 10"-30" 60"-180" 0.5 - 1.0 49.- 3?
1 min 5" - 10" 15" - 20" 01-03 | 02-05"
1s 3-5 3" - 20" 0.01-0.02 .02°-0.05°
Velocity
1h 0.5 - 1.0 m/s 200 - 300 /s
1 min 0.03 - 0.10 m/s 1-2m/s
1s 0.001 - 0.003 m/s 0.002 - 0.005 m/s
Position
1h 500 - 1000 m 200 - 300 km
1 min 0.3-1.0m 30-50m
1s 0.02 - 0.05 m 0.3-0.5m
Tab. 2: Current INS Performance.
4. ACCURACY ACHIEVABLE BY AN INTEGRATED
INS/GPS
As is obvious from Tables 1 and 2, the stand-alone
accuracy of each system will not give the highest
possible accuracy. INS will have superior orientation
accuracy, GPS superior position accuracy. Thus, an
integration of the two systems will result in an optimal
solution which will also provide much needed
redundancy.
GPS positioning using differential carrier phase is
superior in accuracy as long as no cycle slips occur.
GPS relative positions are, therefore, ideally suited as
INS updates and resolve the problem of systematic error
growth in the IMU trajectory. On the other hand, the IMU-
derived attitude is usually superior to that obtained from
a GPS multi-antenna system. In addition, IMU-derived
position differences are very accurate in the short term
and can thus be used to detect and eliminate cycle slips
and to bridge loss of lock periods. Because of the high
data rate, they provide a much smoother interpolation
than GPS. Integration of the two data streams via a
Kalman filter thus provides results which are superior in
accuracy, reliability, and homogeneity.
The following figures will illustrate the orientation
accuracy currently achievable with an integrated
INS/GPS. Position accuracy is not discussed in the same
detail because it is largely dependent on GPS accuracy
for which Table 1 and the references given there can be
consulted.
Fig. 2a shows the attitude output of a navigation-grade
INS in static mode over a period of about 25 minutes.
The systematic error is smooth and reaches a minimum
after about 21 minutes. It is clearly part of the Schuler-
type oscillation. The noise about this trend is very small
and shows white noise characteristics. After eliminating
the trend, the noise pattern in Fig. 2b results. It shows
the system attitude noise which is at the level of 3
arcseconds (RMS) and represents the attitude accuracy
achievable under ideal conditions. Fig. 2c shows the
noise of the same system mounted in an aircraft with the
engines switched on. As before, the trend has been
eliminated. In this case, the low noise level of the lab test
cannot be maintained. The noise is now between 15 and
0.093
0.02} 4
E 0.01} -
"B
Or 4
-0.02 ic =
time (=)
Figure 2a: Total roll error, static case, lab
time (©)
Figure 2b: Roll noise, static case, lab (G =3")
0.02 T T T T 7d, T T T
roll error (deg)
E
Jj
? 1 1 1 1 1 L 1 1
eO 600 700 800 S900 1000 1100 1200 1300 1400
time (sec)
Figure 2c: Roll error, static case, aircraft, engine
switched on (O - 18")
0.02 T T T T T T T T
e
©
Un
T
1
roll error (deq)
C
0020 600 700 E00 S00 1000 1100 1200 1300 1400
time (sec)
Figure 2d: Roll error, static case, aircraft, engine on,
vibration filter (11")
70
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B6. Vienna 1996
Fit
tyr
ck
thi
rel
Sp