13 Tracking of Moving Objects with Accuracy Guarantees
297
In Fig. 13.4, the curve for the constant-speed optimal policy gives the lower
bound for the number of updates needed by the segment-based policy. The deviation of the segment-based policy using the non-modified road network from the
optimal case is substantial. Using the modified road networks, the performance is
significantly closer to the optimal case. For example, for a threshold of 200 m, the
use of the road network modified using the direction-based approach increases the
average time duration in-between consecutive updates from 32 vs. 30 s to 52 vs. 63 s
(for the INFATI vs. AKTA data sets) in comparison to the use of the unmodified road
network.
13.5.2 Practical Tracking Using Routes
We may assume that individuals who travel do so to reach a destination; and the same
routes are often used multiple times. For example, a person going from home to work
may be expected to frequently use the same route. This behavior is confirmed by the
GPS logs available [10, 16].
Taking advantage of knowledge of the routes used by a moving object holds
potential for reducing the number of updates caused by segment changes. Since a
route is a sequence of connected, partial road segments, a route is represented simply
as a polyline. Therefore, segment-based tracking that is applicable to any polylines
is also directly applicable to routes.
All that is needed is to collect the routes of each user. Routes may be obtained
via a navigation system. This case may occur when the user travels in unfamiliar
surroundings. When the user travels in familiar surroundings, it is likely that the user
travels along a route that was used previously. In this case, a system that gathers the
user’s routes and associates usage meta-data (e.g. times of the day and days when
the routes are being used) with these can be used [2]. Such a system is able to return
likely routes on request.
When using segment-based tracking with routes, we effectively assume that we
know the future positions of an object. This is like in the theoretical, constant-speed
optimal policy. The differences are that the polylines that represent routes are created
from the road network, not from GPS logs, and that deviations from the assumed
route are handled. Specifically, if an object deviates from its route, this is treated
simply as a segment change. This will then most likely trigger an update.
When the route of an object is guessed successfully, the theoretical technique
and the practical, segment-based technique have essentially the same performance.
Slight deviations may occur because the routes used by the techniques differ: the
routes used by the theoretical technique are constructed from GPS points and
are more detailed, and slightly longer than those used by the segment-based
technique.
Figure 13.10 illustrates the performance of segment-based tracking using routes
for the INFATI and AKTA data. We conclude that exploiting knowledge of the routes
used by an object can eliminate virtually all updates caused by segment changes that
significantly improve the performance of the segment-based policy.
297
In Fig. 13.4, the curve for the constant-speed optimal policy gives the lower
bound for the number of updates needed by the segment-based policy. The deviation of the segment-based policy using the non-modified road network from the
optimal case is substantial. Using the modified road networks, the performance is
significantly closer to the optimal case. For example, for a threshold of 200 m, the
use of the road network modified using the direction-based approach increases the
average time duration in-between consecutive updates from 32 vs. 30 s to 52 vs. 63 s
(for the INFATI vs. AKTA data sets) in comparison to the use of the unmodified road
network.
13.5.2 Practical Tracking Using Routes
We may assume that individuals who travel do so to reach a destination; and the same
routes are often used multiple times. For example, a person going from home to work
may be expected to frequently use the same route. This behavior is confirmed by the
GPS logs available [10, 16].
Taking advantage of knowledge of the routes used by a moving object holds
potential for reducing the number of updates caused by segment changes. Since a
route is a sequence of connected, partial road segments, a route is represented simply
as a polyline. Therefore, segment-based tracking that is applicable to any polylines
is also directly applicable to routes.
All that is needed is to collect the routes of each user. Routes may be obtained
via a navigation system. This case may occur when the user travels in unfamiliar
surroundings. When the user travels in familiar surroundings, it is likely that the user
travels along a route that was used previously. In this case, a system that gathers the
user’s routes and associates usage meta-data (e.g. times of the day and days when
the routes are being used) with these can be used [2]. Such a system is able to return
likely routes on request.
When using segment-based tracking with routes, we effectively assume that we
know the future positions of an object. This is like in the theoretical, constant-speed
optimal policy. The differences are that the polylines that represent routes are created
from the road network, not from GPS logs, and that deviations from the assumed
route are handled. Specifically, if an object deviates from its route, this is treated
simply as a segment change. This will then most likely trigger an update.
When the route of an object is guessed successfully, the theoretical technique
and the practical, segment-based technique have essentially the same performance.
Slight deviations may occur because the routes used by the techniques differ: the
routes used by the theoretical technique are constructed from GPS points and
are more detailed, and slightly longer than those used by the segment-based
technique.
Figure 13.10 illustrates the performance of segment-based tracking using routes
for the INFATI and AKTA data. We conclude that exploiting knowledge of the routes
used by an object can eliminate virtually all updates caused by segment changes that
significantly improve the performance of the segment-based policy.