For DGPS positioning at the decimeter level, the carrier 
phase observable is required. Once the carrier phase integer 
ambiguities are resolved, the positioning accuracy is at the 
level of a few centimetres to decimeters for separations of 
less than 10 km, depending on multipath and atmospheric 
effects, see Table 4. The tropospheric effect may be rather 
significant for airborne positioning as reported in Tiemeyer 
et al. (1994). As in the smoothed pseudorange case, the 
orbital and atmospheric errors decorrelate with an increase in 
the monitor-remote separation at the level of 2-5 ppm. Shi 
and Cannon (1994) presents airborne results at the few 
decimeter level for baselines up to 200 km when dual 
frequency receivers were used in conjunction with precise 
orbits. 
When using the differential mode of operation (DGPS), 
receiver characteristics become very important in the 
determination of the level of performance. Table 1 shows 
that RMS accuracies of 0.3 - 3 m can be reached for the 
horizontal component while 0.5-4 m can be achieved 
vertically using the smoothed pseudo range DGPS model. 
The range takes into account the noise of the measured 
pseudo range as well the receiver's susceptibility to 
multipath, which are the two dominant errors for a 10 km 
monitor-remote separation. Several tests have confirmed 
values at the high accuracy end of the range when narrow- 
correlator C/A code receivers or P code receivers are used 
(e.g. Cannon et al., 1992). Standard C/A code receiver 
technology typically gives DGPS accuracies at the upper end 
of the range. The accuracy degrades as the separation 
between the monitor and remote receivers increase which is 
due to additional errors from the orbit and atmosphere. All 
results given are for post-mission operation. Real-time 
operation is possible with an appropriate data link but 
accuracies are typically degraded with respect to those given 
in Table 3. 
The use of GPS for attitude determination can take on a 
variety of forms ranging from a self-contained system which 
has a number of channels divided between several antennas, 
to an independent system which is made up of individual 
antenna/receivers. The second option is flexible in the sense 
that the receivers can be used for a variety of applications in 
addition to attitude determination. A minimum of two 
antennas is required for heading determination while at least 
three are needed to obtain roll, pitch and azimuth. Redundant 
antennas are also generally used to improve system 
reliability. The achievable accuracy is mainly a function of 
baseline length between antenna pairs, i.e. the longer the 
separation the higher the accuracy. Table 1 gives the level 
of accuracy as a function of separation which is usually 
determined by platform limitations. Many land-based 
applications require shorter baselines, while airborne and 
marine platforms can tolerate separations of the order of 5 - 
10 m. A number of tests have been conducted with multi- 
antenna systems and accuracies which fall within the levels 
presented in Table 1 have been demonstrated, see e.g. 
Schwarz and El-Mowafy (1992) for a comparison with INS 
and Schade et al. (1993) for a comparison derived from 
inverse photogrammetry. The accuracies given in the table 
may be somewhat conservative. Recently, El-Mowafy and 
Schwarz (1994) reported accuracies of 3 arc minutes (RMS) 
for a 3m baseline. 
5.2 INS Accuracy 
Inertial sensors are devices sensing either linear acceleration 
or angular velocity. When assembled into an inertial 
navigation unit, they instrument an autonomous system for 
three-dimensional velocity, position, and attitude 
determination. In the following, only such systems, not 
individual inertial sensors, will be discussed. 
Inertial navigation systems come in two major varieties, 
namely as stable platform systems and as strapdown 
systems. Stable platform systems establish the internal 
attitude reference mechanically and provide a platform 
orientation with respect to some prescribed coordinate 
system. In strapdown systems, the same process is done 
analytically, i.e. angular velocities with respect to the body 
frame are measured at a high rate and orientation changes 
with respect to the prescribed frame are computed. Although 
both types of systems can be used for the problem at hand, 
strapdown systems have major advantages in terms of data 
rate, attitude output, failure rate, price, power requirements, 
and weight. The following discussion will therefore be 
directed towards systems of this type. 
In georeferencing applications, inertial systems function 
mainly as precise attitude systems and as short-term 
interpolators for velocity, position, and attitude. Because of 
the time dependence of all major errors, the long-term 
velocity and position performance is not sufficient for 
precise georeferencing. Thus, regular position and/or 
velocity updates are needed to keep the overall errors within 
prescribed boundaries. The accuracy of inertial systems 
depends heavily on the quality of the sensors used which 
themselves are a function of the system costs. We will 
therefore distinguish between high accuracy, medium 
accuracy, and low accuracy systems. 
Their error characteristics are summarized in Table 4 using 
four typical time intervals. The one hour interval 
characterizes the long-term behaviour and is therefore an 
indicator of the suitability of using an INS as a stand-alone 
georeferencing system. The one minute interval 
characterizes the short-term interpolation accuracy, 
including bridging for GPS outages, and GPS cycle slip 
detection and fixing. The one second interval characterizes 
the interpolation time for an integrated GPS/INS, assuming 
that gyro drift can be eliminated. The attitude accuracies 
given are for pitch and roll, those for heading are three to 
five times larger. Noise levels in attitude can vary 
considerably, depending on the type of gyros used. In 
general, dithered ring-laser gyros have higher noise levels if 
the compensation loops have not been specifically designed 
for high short-term attitude and velocity. 
Table 4 shows that only high accuracy inertial systems can 
be used for stand-alone georeferencing, and that even in that 
case, the positioning accuracy is marginal. On the other 
hand, in the short term, the attitude accuracy is far superior 
to that obtained from GPS multi-antenna systems. Thus, for 
short-term interpolation of position and attitude, both the 
high accuracy and the medium accuracy inertial system are 
suitable. Both will also provide the velocity accuracies 
needed for motion compensation in SAR systems. It appears 
therefore that an integrated GPS/INS provides the best 
guarantee for a georeferencing system that will cover a wide 
range of applications. 
196 
Error in 
EEE 
Attitude 
lh 
1 min 
ls 
50Hz (nois 
  
Velocity 
Ih 
1 min 
ls 
50 Hz (nois 
  
Position 
1h 
| min 
ls 
50 Hz (nois 
  
  
  
  
——— 
Accu 
Requ 
  
XM 
0.05 - O.1 r 
15" - 30" 
0.0002 - 0. 
  
— 
0.05 - 0.1 r 
  
2-5m 
> 10' 
0.01 - 0.02 
  
2-5m 
2] 
0.0002 - 0. 
  
6. PC 
In Table 5 
accuracy raı 
are listed. | 
also attache 
and dedicat 
in most cas 
and certific