In this section, we present an overview of the BlueDetect system firstly. After that, we introduce the design details of transition methodologies between outdoors/semi-outdoors/indoors and the iBeacon mode of BlueDetect for semi-outdoor environments, respectively.
4.2. Seamless Transition between Outdoors and Semi-Outdoors
For pure outdoor environments, GPS is capable of providing sufficient positioning accuracy. When it comes to semi-outdoor or indoor environments, the number of visible satellites is decreased, and a significant decrease of the GPS signals is expected due to the block of the line-of-sight between the satellites and mobile device. The performance of GPS is jeopardized dramatically while draining the battery at a high power rate under this circumstance. Therefore, BlueDetect will turn off the GPS module in both semi-outdoor and indoor environments.
We conducted experiments to analyze the variation of GPS signals when a client with a mobile device was moving from outdoor to indoor environments (three experiments were conducted in a covered corridor, a connection between buildings and a semi-open parking garage).
Figure 3 illustrates the maximum, mean and minimum of GPS SNR readings from visible satellites in outdoor, semi-outdoor and indoor environments. As shown in
Figure 3, the value of mean SNR dropped more than 20% when coming from the outdoor to the semi-outdoor environment. Clearly, it is more suitable than the other two values to be used as a trigger indicating the switching of environments. Algorithm 1 shows the transition methodology between outdoors and semi-outdoors of BlueDetect.
Algorithm 1 BlueDetect IO detection and localization algorithm (outdoors ⇌ semi-outdoors). |
input: B - Bluetooth signal (iBeacon), G - GPS signal, - switching threshold, - duration threshold output: Location of the mobile device case Outdoor ⇒ Semi-outdoor if for then Turn on Bluetooth; if No less than 2 beacons’ then Turn off GPS; Utilize B for localization; Environment Type ← Semi-outdoor else Turn off Bluetooth; Utilize G for localization; Environment Type ← Outdoor end if end if case Semi-outdoor ⇒ Outdoor if then Turn off Bluetooth, Turn on GPS; Utilize G for localization; Environment Type ← Outdoor else Utilize B for localization; Environment Type ← Semi-outdoor end if
|
If the decline of mean GPS SNRs is detected and the value of the decrease is larger than the switching threshold
for three consecutive samples (sampling rate: 645 ms/sample), the Bluetooth module will be turned on searching for BLE beacons. When no less than two beacons are detected and their RSSs are larger than
= −70 dBm (the detailed methodology of setting
is presented in
Section 4.4), it is confirmed that the mobile device is in the semi-outdoor environment, followed by turning off the GPS to save energy and activating the iBeacon mode of BlueDetect for positioning. Otherwise, Bluetooth will be turned off, and GPS will still be used for localization.
Conversely, when the RSSs of BLE beacons installed at the boundary of the semi-outdoor environment are detected (), the intention of going outdoors can be inferred, and then, the GPS module will be turned on for outdoor LBS in advance for a seamless transition. If no BLE beacon signal is detected for a period of longer than a threshold , which is four consecutive samples (sampling rate: 645 ms/sample) and a relatively stable GPS signal is maintained, the Bluetooth will be switched off.
4.4. Semi-Outdoors (iBeacon)
Reliable LBS in a semi-outdoor environment is not readily available since neither GPS nor IPS can perform satisfactorily in this scenario. We employ the emerging iBeacon technology to fill in this gap. iBeacons make use of BLE proximity sensing to broadcast their unique identifiers to nearby portable mobile devices and trigger a location-based action on these devices. Since the iBeacon protocol uses very short duration messages and does not need a paired connection with mobile devices (broadcast only), it is much more power efficient than classical Bluetooth protocols and less power hungry on the user-side than GPS and WiFi [
34]. With such a merit, a BLE beacon can run on a coin cell battery for months or even for years. According to a recent study on the battery life of 16 major iBeacon hardware devices [
36], by setting the advertising interval as 645 ms, an iBeacon with a CR2450 620-mAh coin cell battery is able to provide 11.2 months of life, which increases to two years as the advertising interval is increased to 900 ms. Nowadays, the iBeacon protocol is becoming a built-in standard for mobile devices, and a high density deployment of BLE beacons in buildings for multiple purposes will be expected in the near future.
For the iBeacon mode of BlueDetect, only a few of the portable, low-cost and battery-powered BLE beacons are deployed as landmarks in semi-outdoor environments for context detection, as well as localization and navigation. From the energy saving perspective, both GPS and WiFi modules of mobile devices are turned off when the semi-outdoor status is confirmed, since they supply no valuable information for IO detection and LBS under this circumstance. The common geographical structure of a semi-outdoor environment, such as a corridor or a sky bridge, is elongated; thus, a sparse deployment of BLE beacons is adequate to cover the entire area. According to the iBeacon protocol, the unique identifying information of each beacon is proximity universally unique identifier (UUID), major value and minor value. These parameters can be used to identify the building, the floor and the exact location of each BLE beacon. With the RSSs and locations of these beacons, we leverage weighted path loss (WPL) [
7], a log-distance path loss model-based localization algorithm, to estimate the real-time location of a client’s mobile device in semi-outdoor environments. The methodology of WPL is described as follows: suppose a client’s mobile device receives RSS from
n BLE beacons simultaneously. The RSS of
i-th iBeacon
can be expressed as:
where the reference path loss coefficient
and the path loss exponent
α need to be calibrated, and
represents a zero Gaussian random noise with standard deviation
σ. Then, based on Equation (1), the distance
between the mobile device and the
i-th iBeacon is calculated by:
The real-time estimated location of the mobile device,
, is computed by:
where
is the normalization constant and
indicates the physical coordinates of the
i-th iBeacon.
In an effort to make the log-distance path loss model robust in a semi-outdoor environment, we first conducted experiments to analyze the effects of Non-line-of-sight (NLOS) and the orientation of mobile device on the RSS from BLE beacons.
We performed an experiment in a covered corridor (typical semi-outdoor environment) to analyze the effects of NLOS firstly. The signal strengths were measured at several different distances away from a beacon, which was attached at the ceiling of the corridor. At each location, one user carried a mobile device (Nexus 6) facing toward the beacon, and 100 RSS samples were collected under the line-of-sight (LOS) condition. In addition, another 100 RSS samples was collected when another occupant was standing between the user and the beacon to block the LOS as an NLOS condition.
Table 2 compares the mean RSS values under both LOS and NLOS conditions at 1–9 meters away from the beacon. As shown in
Table 2, the NLOS RSS value is 4–5 dBm smaller than the LOS RSS value at each location, because the obstacle (occupant) attenuated the signal strengths. Therefore, we further consider the NLOS effects in the process of log-distance path loss modeling for localization in a semi-outdoor environment.
In addition, we conducted an experiment in the covered corridor to evaluate the influence of different orientations of mobile devices on the RSS emitted from Beacons. Similar to the last experiment, we recorded the signal strengths at several different distances away from a beacon placed at the ceiling of corridor. At each location, one user carried a mobile device (Nexus 6) facing four different orientations (
,
,
,
) to measure the RSS values. 100 RSS samples were recorded at each orientation.
Table 3 compares the mean RSS values of four orientations at nine distinct locations from a beacon. It can be observed from
Table 3 that the RSS values of the
holding orientation at all locations are largest among the four directions, because the LOS condition is satisfied. On the other hand, the RSS values of
are smallest, since the user blocked the LOS. The RSS values recorded at the orientations of
and
are usually similar, because their signal transmission conditions are similar. In summary, the average RSS variation caused by different holding orientations of the mobile device is 1.929 dBm, which should not be ignored for log-distance path loss modeling.
Therefore, we include the effects of NLOS and different holding orientations of the mobile device to precisely estimate the parameters
and
α in the log-distance path loss model for semi-outdoor localization. To be specific, we measured RSSs at 14 different distances away from a BLE beacon put at the ceiling of the corridor. At each location, one user carried a mobile device facing four different orientations with distinct LOS conditions, (
, total LOS;
, total NLOS;
and
, partial LOS), to measure the RSS values. 100 RSS samples were recorded at each orientation.
is determined as the mean RSS value at a 0.5-m distance, and
α is estimated by the least-squares method. Based on our experimental results, the estimated values of
is −56.75 dBm and
α is 1.577. The raw RSS measurements and mean RSS values at each reference point are demonstrated in
Figure 4, with the curve fitting by the least-squares method. As shown in
Figure 4, the RSS value decreases to −70 dBm and remains at this level after 6.5 m. Therefore, we define the minimum effective RSS broadcast from each iBeacon to be
= −70 dBm. Accordingly, BLE beacons only need to be placed with 7–10-m intervals for localization in a semi-outdoor environment.