1-2-6 
This information along with the performance achieved in 
different environments is analyzed and then stored in the 
expert system database. The information in the database is 
then used to processing strategy for the real-time and post 
processing modules. 
4. BUILDING THE EXPERT KNOWLEDGE 
DATABASE 
Since this accuracy can always be achieved for good 
satellite coverage and signal reception, expert knowledge is 
in this case mainly concerned with countermeasures to 
cases of poor GPS satellite geometry, signal blockage, or 
cycle slips. INS aiding of GPS plays a major role in all of 
these countermeasures. Since they have to take effect 
before GPS positioning accuracy deteriorates beyond half a 
cycle, a real-time system for the detection of such 
situations is needed to alert the operator. In addition, expert 
knowledge on INS bridging and backward smoothing are 
important parameters for the real-time decision. Such 
knowledge can then be transferred to the knowledge 
database then to the real-time component to control the 
ZUPT time, for more details see Schwarz et. al., 1994. 
Figure 6 depicts the error behavior of the INS in stand 
alone mode. GPS observations were removed from the data 
for a 60 second period and the INS data were processed in 
stand-alone mode. The truth model of this diagram was 
obtained by using the trajectory computed from the original 
GPS/INS measurements. Its accuracy is good to a few 
centimeters. The INS stand-alone positioning results stay 
below the half cycle level (10 cm for LI and 43 cm for 
wide lane) for about 30 seconds for LI and 65 seconds for 
the wide lane. 
0.5 
P 0 
-1.5 
- 
—i—i—i—i—i—i—i—i—r—i—i—1 
1 
’ 
21 
41 61 
Time (sec) 
81 101 
Figure 6: INS Error Behavior in Stand-alone Mode 
The output from the Kalman filter was processed again by 
the Rauch-Tung-Striebel (RTS) optimal smoother, for more 
details about the RTS optimal smoother, see Gelb (1979). 
The results of the smoothing are plotted in Figure 7, it has 
been assumed that there is a GPS position update at the end 
of the 100 sec. The truth model of this diagram was 
obtained by using the trajectory computed from the original 
GPS/INS measurements. They show that the smoother has 
improved the accuracy and that the INS bridging interval is 
extended to 100 seconds for both LI and wide-lane case. 
Figure 7: INS Error Behavior in Stand-alone Mode after 
Backward Smoothing 
Signal blockages through houses and trees or complete loss 
of lock are indicated by the receiver hardware. Thus, real 
time detection of these events is not a problem and an alert 
can always be given. Fixing ambiguities afterwards is an 
involved process that depends very much on the specific 
situation. Countermeasures are INS bridging and 
smoothing, as well as ambiguity resolution on the fly or a 
combination of all of these. Such information can easily be 
monitored in each survey and stored in the expert system 
database. Figure 8 shows the expert system database for the 
OTF time to fix the ambiguity for a number of surveys. The 
table shows that at present only sparse information is 
available. It will get more fully populated as information 
from different surveys will increases the knowledge 
database. 
Knowledge View 
1 \ 
Initial Static I GPS Static 0n-Ihe-Fljj Production j 
■ -, # 
L ■ ■ • 
Session ¡VA000614 3 ** Wide Lane 
r L1/L2 
m3 
ÇJew! 
4 
Number of Satellites 
5 6 7 
8 
500 m H| 
48 
25 
25 
21 
59 
67 
28 
5.0 km ÎH 
69 
65 
48 
53 
7.5 km I 
81 
79 
10 km ^ 
162 
120 
71 
1.1 km 
239 
15 km |*j| 
186 
?liknir 
227 
■20krnB 
356 
OK Cancel 
Figure 8: The VIS AT Expert System Database 
CONCLUSIONS 
In this paper, the development of an expert knowledge base 
system for MMS systems has been presented with emphasis 
on systems using GPS, INS, and imaging sensors. The 
expert knowledge is contained in the calibration, planning, 
survey mission, and real-time and post-mission quality 
control. The expert system encompasses the total of this