Integration of GPS with Other Sensors and Network Assistance - Understanding GPS: Principles and Applications - page 514

try of the satellites (i.e., none are likely to be visible below the local horizon),
GPS-based altitude estimates are less accurate than lateral position estimates, so an
augmentation for improved altitude estimation is attractive. Relatively low-cost
barometric altimeters are available [51], which can sense changes as low as 1m in
altitude and, with proper calibration, can provide absolute altitude measurements as
good as 10m. Both the absolute and relative altitude information are valuable in
aiding GPS, whether or not inertially augmented.
Since any barometric altimeter determines altitude through sensing air pressure,
calibration (which associates pressure readings and altitudes) is necessary, and the
calibration information will degrade relatively slowly with time, as local weather
conditions change. The calibration accuracy will also degrade as the physical sepa-
ration between the vehicle and the reference location for the calibration increases.
Thus, operation of a vehicle navigation system that makes use of a barometric altim-
eter as a source of absolute altitude information will require either that calibration
data from a reference station be supplied to it or that similar calibration information
be supplied by GPS. If the navigation system is integrated with a cellular phone or
other communication means within the car, then the communications network can
provide the calibration information, if located relatively close to the car (most cellu-
lar phone service areas are limited to mobile separations less than 10 km). Air pres-
sure sensors can be implemented on a small silicon chip at low cost [51].
Unfortunately, at the time of this writing, cellular phone base stations (BSs) did not
support such a capability, although proposals have been made fairly recently [52];
thus, calibration using GPS will be required. This is preferably done by inclusion of a
barometric altimeter bias state in the Kalman filter, which compares GPS altitudes
(when relatively accurate altitudes are available with the barometric altimeter read-
ing) with barometric altimeter readings: an appropriate level of process noise associ-
ated with the barometric altimeter bias will ensure that the calibration is not static.
In addition, if the barometric altimeter-derived pressure changes are used as a source
of vertical velocity information by the integration filter, a scale factor error state that
calibrates altitude change derived by pressure change using GPS-determined vertical
velocity may be necessary.
9.3.2.6 Magnetic Compass
Magnetic compasses provide an inexpensive means of determining vehicle heading
and have been used to augment early DR systems [53]. The major problem associ-
ated with the use of a magnetic compass as a primary or sole heading reference is its
sensitivity to magnetic anomalies. Although compass designs can be self-calibrating,
this calibration serves only to remove the static disturbance of the Earth's magnetic
field (e.g., as could be induced by the vehicle itself). The error induced by the tilt of
the sensor can also be compensated [29]. Dynamic sources of disturbance, which
could be generated by other passing cars or the steel trusses of a bridge, can induce
very significant errors in the compass' heading indication. Thus, the compass is usu-
ally relegated to a backup role or as a complement to another system. If integrated
with a source of heading rate information (e.g., as could be supplied by a low-cost
gyro or an ABS), the integration filter's residual test, which compares the current
magnetic compass reading with the current best estimate of heading (propagated