International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
Accelerometer Scaling Factor RMS XYZ: PPM 
  
  
  
  
  
    
  
  
  
05 T T T T T T I 
; : ; PRET 
: = RMS Y 
: RMS Z 
0 Eu exelente leeren TEE E 
Toe, ; 
dh da hey mae ran ser warren N 
os XE 1. ee Een 7 
: : ceci GE 
à NB aT BE Te 
ji f: 
Adee : y uomen i ip fs MST 5 
e; | 
i5r--—— ft ere forint y ES EE ue E 
3 i i i i i i i 
0 500 1000 1500 2000 2500 3000 3500 4000 
. Start Epoch: 322970 End Epoch: 326799 
Figure 8. Difference in RMS of accelerometer SF 
corresponding to Figure 7; negative sign indicates 
improvement due to DOV compensation. 
Further improvement in the INS/DOV solution can be 
achieved by applying periodic ZUPTs. ZUPT seems to have 
relatively more impact on the position coordinates, as 
compared to the attitude angles. The calibration performed 
during ZUPT also affects the INS/DOV trajectory portion 
that follows the ZUPT event, and thus, the difference 
between the two solutions still exists even though both 
solutions are based on INS/DOV only after the ZUPT event. 
For more details on ZUPT effects on INS navigation 
accuracy, see Grejner-Brzezinska et al. (2001). The effects of 
using DOVs in the navigation algorithm and performing 
ZUPTs to calibrate the (observable) errors can be clearly 
seen by comparing the reference GPS/INS/DOV solution 
under a favorable GPS constellation with the corresponding 
INS/DOV/ZUPT solution. The INS/DOV/ZUPT solution 
was calibrated by GPS prior to turning off the GPS signal; no 
ZUPTs were performed for the GPS/INS/DOV solution, even 
for the static portion of the trajectory. Our tests indicate that 
the free INS solution supported by DOVs and ZUPTS is 
capable of providing horizontal coordinates within an 
absolute difference of 1-3 cm (Figures 11-12), as compared 
to the reference “truth” (GPS/INS/DOV), while the attitude 
angles compare at a 1 arcsec level (Table 3). More details on 
the impact of the DOV compensation on the navigation 
solution can be found in Grejner-Brzezinska et al. (2003). 
3. SUMMARY AND CONCLUSIONS 
Based on the analyses summarized in this paper it can be 
concluded that the attitude components (primarily pitch and 
roll) are more affected by DOVs than the position 
coordinates, and the effect is more pronounced during the 
loss of GPS lock. The combined effects of DOVs and ZUPTS 
were analyzed for the land-based data set, indicating that 
while DOVs influence primarily the attitude, the ZUPTs 
have more impact on the position solution. It was also 
demonstrated that the use of DOVs and ZUPT calibration 
during the loss of GPS lock is capable of bringing the 
combined solution to accuracy comparable with the reference 
GPS/INS/DOV solution. The data sets used here were 
collected in test areas with relatively small DOVs; still their 
effect on the sensor errors and ultimately on the position and 
attitude solutions is visible. More tests are needed in areas 
with larger DOV magnitude and variation. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Component | Mean | Std | Max | Min | Units 
120-second test duration, 1300 s prior calibration 
RMS N -10 11 0 -39 mm 
RMS E -]1 j^ 0 -41 mm 
RMS Ht 0 0 0 0 mm 
RMS Vn 0 0 0 -Î mm/s 
RMS Ve 0 0 0 i mm/s | 
RMS Vd 0 0 0 0 mm/s 
RMS Head. 0 0 0 0 arcsec 
RMS Pitch -4 0 -4 -4 arcsec 
RMS Roll -4 0 -4 -4 arcsec 
360-second test duration, 1300 s prior calibration 
RMS N -214 230 0 -793 mm 
RMS E -218 234 0 -806 mm 
RMS Ht 0 0 0 -] mm 
RMS Vn -2 2 0 -6 mm/s 
RMS Ve -2 2 0 -6 mm/s 
RMS Vd 0 0 0 0 mm/s 
RMS Head. 0 0 0 0 arcsec 
RMS Pitch -4 0 -3 -4 arcsec 
RMS Roll -4 0 -4 -4 arcsec 
  
  
  
  
  
  
  
Table 2. Position and attitude accuracy improvement 
between the INS/DOV and INS-only solutions, summary of 
statistics; land-based test. 
  
  
  
  
  
  
  
  
Difference Mean Std Max Min Units 
North 13 31 55 -15 mm 
East 16 15 38 -13 mm 
Height -26 AS 67 -101 mm 
Heading 0 0 1 -1 arcsec 
Pitch 0 0 1 0 arcsec 
Roll 0 0 1 -] arcsec 
  
  
  
  
  
  
  
Table 3. Position and attitude difference between solutions 
(5) and (4); summary of statistics (land-based test); 
  
  
  
  
85-s ZUPT. 
North, East, Height Difference of INS--DOV and INS alone: Meter 
0.08 T T T 
: ' : : — North 
; : : «en East : 
0.07 LS TRIE Height aie / nea i 
Meter 
  
  
  
  
0 20 40 60 80 100 120 
? Epoch Fram 324269 To 324369 [Second] 
Figure 9. Coordinate difference between INS/DOV and INS- 
only solutions after ~1300 s of GPS-based calibration; 
land-based test. 
      
   
    
  
   
  
   
   
    
   
   
    
ArcSec 
Meter 
Mater