Integration of GPS with Other Sensors and Network Assistance - Understanding GPS: Principles and Applications - page 506

bearing within a tolerance of the vehicle heading. When the vehicle makes a turn, the
list of candidate segments is further reduced based on examining the topology of the
road network to find candidate segments that have a turn in the direction the vehicle
turned. Through this process, the list of possible vehicle positions is eventually
reduced to a unique segment, and the confidence in the positioning solution
increases accordingly. When there is only one possible vehicle position, the
map-matched position solution will have a small confidence region and therefore
can be considered highly reliable. If a position jump or a turn occurs that introduces
additional potential positions in the road network, then the confidence region
should grow, reflecting a lower confidence in the map-matched solution.
In order to support map matching, the map data should have high position
accuracy, ideally better than 15m, to minimize incorrect road selections. The map
data should also be topologically correct, reflecting the real-world road network,
so that the algorithm does not get confused if the user drives on a road that is not in
the database. The expected accuracy of the road centerline data should be used
in the map-matching process to determine the overall confidence region of the
map-matched position solution.
Once a match is determined, the vehicle position is then displayed on the
matched road segment and used for the route guidance instructions. The map can
also be used as a sensor itself to provide useful information to the positioning sub-
system or to calibrate inertial and other DR sensors. These capabilities are broadly
referred to as map aiding [47] and map calibration.
Map aiding is most useful when map matching has determined that the vehicle
has just turned a corner, in which case its position is in close proximity to the inter-
section of two streets of a known location in the map database. This reference posi-
tion may be treated as a single position fix by the integration filter (see Section 9.3.3
for further discussion), which serves to correct or improve the accuracy of the abso-
lute position determined by GPS. Further, if map matching has determined, with
high probability, that the vehicle is traveling on a specific road in the database and
that road is straight, then a heading fix may be generated for the integration filter
(see Section 9.3.3) based on the bearing of that road segment according to the map
database, as shown in Figure 9.27. Another way to utilize the heading information
after a turn is to impose a constraint on the model to force the heading of the vehicle
to match the bearing of the road. Map feedback can be used instead of DR sensors to
improve the performance of GPS in low-cost navigation systems [47].
In addition to the horizontal position components of road vectors, ground eleva-
tion data may be used to augment the performance of GPS. A digital terrain model
(DTM) is a representation of the Earth's surface that can be used to extract elevation
data. A digital elevation model (DEM) is a type of DTM with a regularly spaced grid
of elevations corresponding to the elevation of the Earth's terrain at that point.
Modern DTMs are derived from airborne or satellite-based remote sensors, are
georeferenced using GPS coordinates, and have vertical accuracies better than 10m.
Terrain elevation can be used to improve the accuracy associated with GPS fixes
for land applications. As is well known, and addressed previously in the text, the
vertical axis is the weakest part of the GPS solution. Terrain elevation data, if suffi-
ciently accurate, can be added as a constraint to a least squares or WLS GPS fix or
added as a measurement to a real-time Kalman filter. To apply a height constraint,