transmit their ID using a slow radio encoding that takes as long as ultrasound takes to propagate 10 m.
This ensures that the radio transmission from each cricket beacon fully overlaps the flight of the
ultrasound pulse from the same source (illustrated in Figure 3.6). This overlap guarantees that at the
time a listener receives an ultrasonic pulse from a beacon, it will still be receiving that beacons ID
via radio. This allows the listeners to detect (and ignore) the RF collision that would occur if two
or more ultrasound packets were in flight at the same time. This ensures that pulses will never be
incorrectly associated with the wrong beacon ID. To reduce the chance that these collisions occur,
Cricket beacons randomize the time between transmissions. (In their experiments, beacons waited
between 150 and 350 ms between broadcasts.)
Thus far, the systems presented in this chapter estimate the room or portion of a room where the
device is located. The final two systems we present estimate with high accuracy the absolute location
of a device within a room (e.g., “110 cm south, 50 cm east, and 35 cm up from the northwest corner
of Conference Room 206”). The first is the Active Bat system, and it was developed by AT&T Re-
search in 1999. The Active Bat system uses ultrasound receivers to localize small (35 g) pager-like
devices called Bats (shown in Figure 3.7) in three-dimensional space [66]. The system is extremely
accurate, estimating location correctly within 5 cm 50% of the time and within 9 cm 95% of the
time. Like Crickets, Bats are located by measuring the travel time of an ultrasonic pulse. In contrast
to Crickets, however, the Bats emit the ultrasonic pulses, and receivers in the infrastructure listen
for and time the pulse’s travel.
FIGURE 3.7: An AT&T Active Bat [66]. © 2002, ACM, Inc. Reprinted by permission.
The accuracy of the Active Bat system derives from its dense deployment of ultrasound re-
ceiver units (shown in Figure 3.8). In the original AT&T deployment, 100 ceiling-mounted receiv-
ers were used in a 100-m
office space. This dense receiver deployment makes it likely that a Bats
pulse would be heard by at least three and likely many more receivers. To estimate location, receivers
in the Active bat systems pool their time-of-flight observations to a central server that performs
multilateration (a generalization of trilateration) in which location is estimated from three or more
ranging estimates from known locations [66]. As in the Cricket system, reflections of ultrasound are
common, and the Bat system uses the redundancy of the grid of receivers to reject these reflections
as outliers. The success of their approach can be seen in the error CDF shown in Figure 3.9.
The Active Bat system is sufficiently accurate that it is possible to place multiple Bats on a
rigid object and, by measuring the Bats location, reconstruct not only the objects location but also
its orientation. Experiments in which two Bats were attached to a rigid object 22 cm apart showed
that orientation could be estimated within 10° over 80% of the time [66].
The accuracy of the Bat system is dependent not only on the density of ultrasound receivers
but also on centimeter-accurate knowledge of each receiver’s location. For the University of Cam-
bridge’s Bat deployment, this was done with the aid of professional surveying equipment. However,
Scott and Hazas showed that it is possible for systems like an Active Bat deployment to be self-
surveying [151]. They placed a 1 × 1-m wooden frame fit with 21 Bats in a known geometry in a room
and collected 20 min of raw time-of-flight measurements between the Bats and the receivers. By
FIGURE 3.8: Ultrasound receiver unit for Active Bats. Shown sitting on ceiling tile (left) and shown
when installed (right) with only receive sensor exposed [66]. © 2002, ACM, Inc. Reprinted by

Get Location Systems now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.