Integration of GPS with Other Sensors and Network Assistance - Understanding GPS: Principles and Applications - page 509

navigation purposes, producing cross-track error that is more than 15% of distance
traveled. Of course, the error persists only as long the vehicle is turning. If the lever
arm L can be measured, the error effect can be compensated. As a minimum, how-
ever, the real-time navigation filter should recognize this error effect in its weighting
of GPS headings in turns.
As efforts continue to lower acquisition and tracking thresholds for GPS receiv-
ers, additional sources of error must be considered, including false signal acquisi-
tions and tracking of reflected signals (commonly referred to as multipath). As
discussed in Section 5.13, acquisition of signals below normal thresholds requires
longer coherent and noncoherent integration times. As SNR thresholds for acquisi-
tion and tracking are lowered by more than 20 dB, the potential for cross-
correlation (i.e., declaring detection of a higher power signal with an incorrect PRN
code) increases. In addition, the conservatism associated with normal detection
thresholds (i.e., the threshold placed upon the peak-to-noise floor ratio) may be
relaxed in order to increase coverage. Alternate tests may also be employed (e.g., use
of a neighbor test , where a detection may be declared if the peak magnitude and the
next largest peak magnitude are in neighboring code phase positions—separated by
one-half chip). Such relaxations of conservatism in detection inevitably bring a
higher probability of false signal acquisition (i.e., interpreting integrated noise as a
signal). Both false signal acquisition and cross-correlation will produce
pseudoranges that are grossly in error; generally, these errors do not persist as the
transition is made to tracking the signal. If this should happen for a short period of
time, however, statistical rejection tests employed by the navigation filter should
remove them.
Reflected signal tracking is a serious problem that can arise in urban canyons. It
can occur when the direct signal path is obscured by a high-rise building, yet a
reflected signal path is visible to the GPS receiver. Note that this condition cannot
be labeled as multipath, as the direct path cannot be seen, and only the reflected ver-
sion is tracked. Of course, the reflected signal will be attenuated relative to the
direct path, and the geometry of the reflection cannot persist indefinitely.
Pseudoranges presented to the navigation filter will have additional, unexpected
error due to the additional range delay associated with the reflected path, and
Doppler measurements derived from the reflected signal can be significantly in
error. In fact, the measured Doppler component due to receiver motion may be
opposite in sign to the actual Doppler component induced by the receiver motion. It
is generally a function of the velocity of the vehicle relative to the surface that is sup-
porting the reflection, as illustrated in Figure 9.29. In Figure 9.29, as was the case
with the false acquisitions, we must rely upon the integration filter's statistical
rejections to preserve acceptable navigation performance.
9.3.2.4 Transmission and Wheel Sensors
The use of elapsed distance traveled information available in the vehicle is generally
a low-cost, high-value augmentation of GPS. Vehicle transmission and wheel sen-
sors can be used to determine the speed and heading changes of the vehicle. Depend-
ing on the type of sensor utilized, the distance determination can become unreliable
at low speed; if variable reluctance sensing [48] is used, the sensor output becomes