Istanbul 2004 
  
  
  
  
  
  
3500 4000 
estimated by 
solution 3); 
> between the 
| of ~8 arcsec. 
different from 
absorb some 
vector. This 
may remove 
n gyro and 
nponents are 
2 differences 
o solutions (3 
t direct DOV 
. Figures 5-6, 
and its RMS 
'ompensation, 
between the 
estimated by 
ded. Figure 8 
nt due to the 
  
Van 
  
Sern 
10 4000 4500 
e estimated 
n the gravity 
nd (5); 
   
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
TN Change 
wrt 
Error Difference Mean | Std Units no DOV 
comp. 
(79) 
AccBias X -0.0 0:5 0.0 
AccBias Y -0.0 0.4 ug 0.1 
AccBias Z -].1 0.4 34.4 
| AccSF X is T4 Hi 
AccSF Y -0.4 2.1 ppm 32 
AccSF Z 17 0.7 6.8 
GyroBias X 0.6 2.0] arcsec 0.4 
GyroBias Y 4.2 5.2 per 7.3 
GyroBias Z -12.0 6.0 hour 27 
GravErr N 7.4 2.8 103.8 
GravErr E -7.9 2.3 arcsec 56.9 
GravErr D 0.6 0.2 17.2 
RMS Difference … = E 
AccBias X -0.2 0.1 7.8 
AccBias Y -0.1 0.0 ug 7.6 
AccBias Z -0.0 0.0 0.2 
AccSF X -1.4 0.3 62 
AccSF Y -0.8 0.2 ppm 3.5 
AccSF Z -0.3 0.0 13 
GyroBias X -0.4 0.4 | arcsec 3.0 
GyroBias Y -0.5 0.5 per 3.6 
GyroBias Z -0.4 0.2 hour 1.2 
GravErr N -4.8 0.1 100.0 
GravErr E -4.4 0.2 arcsec 100.0 
GravErr D -0.1 0.0 3.3 
  
  
Table 1. Error estimation difference between solutions (5) 
and (3) (the difference in DOVs is between the true DOVs 
and DOVs estimated in solution (3)). 
2.4. DOV effects on navigation solution 
A comparison of the GPS/INS solution enhanced by the 
DOV data with the GPS/INS-only solution indicates that the 
effect of DOVs on position estimates is below a cm, while 
pitch and roll are improved (-4 arcsec RMS improvement) 
by applying DOV compensation. More visible effects of 
using DOVs can be observed by comparing the free- 
navigation INS solution with the INS solution supported by 
DOVs, as presented in Figures 9-10. Clearly, height and 
heading are not visibly affected by external DOV 
information. To fully assess the impact of DOV 
compensation on free-INS navigation, we compared 
solutions with different durations of GPS signal blockage (30 
to 360 s) and different times of INS calibration with the GPS 
signal before loss of lock (—1300 s and — 460 s calibrations 
were considered). As illustrated in Table 2, a longer 
calibration period prior to a GPS gap may contribute to a 
slower error growth during the gap, especially for longer 
gaps; also, the effects of DOV compensation become more 
visible for longer GPS gaps (30 and 60-s gaps were tested 
but they show no visible impact on the results, and are not 
included in Table 2). For example, a 120 s gap with a 1300 s 
prior calibration results in cm-level improvement in 
horizontal coordinate RMS, while a gap of 360 s shows an ~ 
21 cm improvement. Heading does not seem affected, and 
pitch and roll improvement is at the level of ~4 arcsec for all 
cases. 
Accelerometer Scaling Factor X,Y .Z PPM 
en 1 I 1 | | 
  
  
  
  
  
  
l 1 
“s00 ‘000 1500 2000 2500 3000 3500 4000 
Start Epoch: 322970 End Epoch; 326793 
Figure 5. Accelerometer scale factor estimated by GPS/INS 
filter; no DOV compensation. 
Accelerometer Scaling Factor RMS X,Y.Z: PPM 
  
  
5 45 T T T T 
1 1 ; ; : | — RMS X 
—- RMSY 
an oes RMS = 
  
  
  
  
  
  
  
I 
5800 1000 1500 2000 2500 3000 3500 4000 
Start Epoch: 322970 End Epoch: 326799 
Figure 6. RMS for accelerometer scaling factor (SF) from 
Figure 5. 
Accelerometer Scaling Factor X, Y Z: PPM 
  
  
p^-------t-------.-.---e------—Mu.—uui--------4------- 
  
  
  
    
PPM 
   
| 
: : : ; ' 
1 i Ex 
epit anta ttn wen ; hr 
jt ' 
sem Saee imi E-- 
  
  
215 i { { i i i | 
0 500 1000 1500 2000 2500 3000 3500 4000 
Start Epoch: 322570 End Epoch: 326799 
  
Figure 7. Difference in accelerometer SF between solutions 
with and without DOV compensation.