Backfire 2.4 Ghz WiFi Antenna Experiment Settings

The specifications of hardware in our experiment are as
follows. Each directional antenna is a Backfire 2.4 Ghz WiFi
antenna from RadioLabs [11], which has 15 dB gain and 32o
half-power beamwidth. Each antenna is connected to a DLink
DI-524 wireless router [12]. Each pair of antenna and
router is mounted to a manual tripod with 5o angular scale
marker. The example of one antenna is shown in Figure
4(b). All antennas obtain their geographic locations from
landmark mapping using Google Maps service [9]. The
antenna alignment is calibrated by an electronic compass.
We use Nokia N95 phone with built-in 802.11g wireless
interface as the target device in the experiment.
We conduct the experiment on a single-story house shown
in Figure 4(a). The three star-shape points denote the deployment
locations of three antennas. The other 21 circular
points denote the placement of the target devices to collect
the wireless radio signal inside the house. We deploy two
antennas on the sidewalk outside the house and one antenna
in the backyard. Each antenna is set up within a few minutes.
We divide the area into grid with cell size 0.1×0.1 meter2
to balance the trade-off between accuracy and computation
delay. The signal is collected from 21 points. Each directional
antenna has the angle granularity of 5o, which results
in 72 directions (i.e., i = [0o, 5o, 10o, ..., 350o, 355o] for
all i). At each collection point, the wireless radio signal
from each outdoor directional antenna is sampled in one
complete round (72 readings from each antenna) with 15-
second duration for each angle. Due to the device’s reading
delay, 15 seconds is the shortest time for the device to sense
a change of signal level. In reality, we expect the reading
delay to be much faster due to better hardware.
In Section V-B, the localization results at each of 21
collected points are shown. In Section V-C, the accuracy
of the tracking algorithm is evaluated via simulations based
on the trajectory shown in Figure 4(c). In the simulation,
the RSSI at any point is linearly interpolated from the two
closest collected sample points. As the distance between any
two adjacent points in the trajectory is less than 3 meters,
we believe that the interpolated RSSI is realistic enough to
represent the real RSSI at any point in the trajectory.

Localization Accuracy based on real data collection

This section presents the accuracy of the localization
algorithm based on real data collection previously described.
We use the term ’rssi’ mode to refer to localization with the
pure RSSI-based probability estimator (i.e. Section III-B1)
and the term ’rank’ mode to refer to localization with the
correlative ranking-based probability estimator (i.e. Section
III-B2). Unless specifically stated, each localization result
is done with 3 antennas, non-linear stretching parameter
K = 10 and antenna beamwidth offset B = 40o.
Angular Prediction Error: Figure 5(a) shows the distributions
of angular errors of the two probability estimators.
The angular error is the difference between the predicted
AoA and the actual AoA from the target position to the
antenna. The result shows that the ranking-based estimator
is generally more accurate than the RSSI-based estimator
(i.e. 5o VS 10o median error). Note that more than 80% of
errors are within half of the antenna’s half-power beamwidth
(i.e., 16o), indicating that the AoA prediction from both
estimators are good with respect to the hardware capability.
Positioning Error: Figure 5(b) presents the distribution
of positioning errors in meters of the two probability estimators,
with and without the RSSI-distance bound (from
Section III-C3). The results indicate that the rank-based estimator
performs slightly better than the rssi-based estimator.
Also, the use of RSSI-distance bound noticeably improve
the accuracy of the two estimators by more than 50%.
RSSI-distance Bound: To validate the correctness of the
RSSI-distance bound, Figure 5(c) shows the relationship
between the distance from each sample point to each anchor,
and the corresponding RSSI from the strongest direction.
The bound line is the distance upper bound (or RSSI lower
bound) estimation from Figure 3. The results show that
around 90% of the points are within the estimated bound.
Only 10% of the points receive stronger signal than the
bound. One explanation is that these points are focal points
where signal from multiple directions accumulate, resulting
in stronger signal than usual. A better bound for such
uncommon corner cases is left as future work.
Effect of Beamwidth Offset Parameter (Bi): As described
in Section III-C2, the beamwidth offset parameter Bi for
each antenna ai can be used to reduce the positioning error
caused by the antenna’s half-power beamwidth. Figure 5(d)
shows the average positioning errors of the two estimators,
with and without distance bound, for different values of the
beamwidth offset parameter Bi. Since all 3 antennas are
of the same model, we use the same Bi for all antennas.
The results show that the optimal Bi value is around 25o −
35o, which is consistent to the 33o half-power beamwidth
from the antenna’s specification. Hence, we conclude that
Bi parameter for antenna ai can be set to the half-power
beamwidth specification of the antenna ai for good accuracy.
Effect of Non-linear Stretching Parameter (K): Figure 5(e)
presents the average localization errors with different nonlinear
stretching parameter values (K) proposed in Section
III-C1. The results show that the localization performs better
as K grows larger, with the optimal K = 20. This validates
the hypothesis that the AoA-based probability is not linearly
proportional to RSSI value nor RSSI ranking. Note that
when K = 1, the RSSI-based estimator without RSSIdistance
bound exactly represents the work of Niculescu and
Nath [7], which gives the average error of 4.07 meters in
our experiment. By using the ranking-based estimator with
RSSI-distance bound and a proper value of K, we can reduce
the average error down to 2.56 meters.
Effect of Additional Antennas : Figure (5(f)) shows the localization
error distributions of the ranking-based estimator
with RSSI-distance bound when using 2 anchors (i.e., the
two leftmost anchors in Figure 4(a)) and 3 anchors (i.e., all
anchors in Figure 4(a)). The result shows that the accuracy
slightly improves when the number of anchors increases
from 2 to 3. Due to time and resource limitation, we do
not experiment with more anchors. However, we expect the
number of anchors to be less than 4 or 5 due to the quick
set-up time requirement. The question of how to find the
best locations to deploy each antenna is left as future work.

Indoor Tracking in Emergency Scenarios

Many indoor tracking schemes for emergency scenarios have been proposed so far [4], [5], [13]–[16]. Overall, there are three approaches for tracking target inside buildings.

The first approach, a scattered sensor-based technique, predeploys sensors equipped with wireless devices such as RFID and WiFi inside buildings [13], [14]. Although this technique can be used to collect other information than signal strength (i.e., temperature), it is expensive and sometimes
not possible to deploy many sensors beforehand.

On the other hand, the second approach, called dead reckoning [4], [15], exploits inertial sensors and pedometers carried by the targets to estimate moving distance and directions. The accuracy of this approach depends on the accuracy of the inertial sensors. This approach is orthogonal to our work and the combination of both works is possible.

In the third approach, a small number of base stations which emit beacons are set outside the building. Our work is categorized in this approach. However, our work is the first work to use WiFi directional antenna to track the target.

Other candidates include UWB and omni-directional WiFi [5], [16], which usually assume the environment knowledge of the building such as signal fingerprint or per-building signal-distance mapping. Our approach does not assume such knowledge to be known prior to the operation.

WiFi-based Indoor Localization

Works on WiFi-based indoor localization can be divided into two categories. The first category localizes targets by using the received signal strength indicator (RSSI) from omnidirectional antennas to determine the target’s position [1], [2]. This class of localization algorithms assume a mapping between RSSI and spatial metrics (e.g., relative distance or absolute position). To get accurate results, such mapping
must be pre-calculated specifically for each building, which is unsuitable for emergency scenarios.

Another category of WiFi-based localization techniques is to estimate the angle-of-arrival (AoA) and distance from directional antennas to the target location [6]–[8]. Most of works in this category assume several directional antennas placed inside the building to localize a static indoor object.
We believe our paper is the first work to use few outdoor directional antennas to track a moving indoor target.

Choosing the Suitable GPS Tracking Device

Technology is providing various solutions for solving problems as well as to simplify various complications; the invention of Global Positioning System that was made based on the satellite navigation technology is offering the simpler solution to spot and identify a location. It can be very useful for navigation assistance; the use of gps tracking device is now quite common to determine or identify a location and also added in various devices and items in order to regularly record the position. The first project to develop the GPS navigation system was initiated in 1973 to improve the navigation system with all of its limitation at that time.

5

The Common Uses of GPS Tracking Device

The gps tracking device in the present time is commonly used to determine and identify a location with precise information of a thing; the device is also commonly attached to moving object in order to observe and record the movement of the object. The recorded location can be transmitted to a recorder that writes the information to a database or the tracking unit can record the information. The tracking device is also commonly used to track the presence or location of a person; it is a possible solution for the parents to track their child or children’s location.

Things to Consider in Choosing GPS Tracking Device

There are several things to consider for anyone who wanted to purchase gps tracking device in order to get the suitable device with the planned usage or requirements. One of the most important things is the reporting needs; the interval of the recorded location is one of the considerations in choosing the suitable tracking device. The battery capacity is another important consideration; it will affect how often the device require electricity charging while the tracking software is another thing consider in order to get all of the requirements covered.