This paper studies the accuracy improvement of the localization technique based on the received signal strength indicator (RSSI)ranging using Bluetooth, which has been commonly used for an indoor localization. In our system, a transmitter diversity technique is introduced, where a beacon transmits four di erent ID signals with four di erent radiation patterns. A smartphone is used as a receiver, and detects the signal with the strongest RSSI among all beacons. The most likelihood detection is used to improve the localization accuracy, where the close beacon is trusted more than the distant one. The experiment revealed that our diversity scheme improves the average and median localization errors by 1.22 m and 1.86 m, respectively. The most likelihood detection considering the distance realizes 6.70 m and 4.62 m accuracy in average and median values, respectively.
The indoor localization techniques using Bluetooth Low Energy (BLE)
beacons have been widely studied [
In this research, we propose a method to improve localization accuracy by introducing RSSI-ranging with a pattern diversity and distance-based reliability function technique. The experiment in the indoor environment was carried out to verify the accuracy of the proposed method, and its performance compared to the conventional methods are discussed. 2
RSSI-Ranging Method Using Pattern Diversity and Distance-Based Reliability Function In this section, we will show the method to improve the localization accuracy using RSSI-ranging with pattern diversity and distance-based reliability function. RSSI-ranging is performed using received signals transmitted from multiple beacons. However, the estimated distance for each beacon is not accurate due to the multipath fading. The pattern diversity is introduced to improve the ranging accuracy by alleviating the multipath e ect. The reliability function is used to take multiple ranging results into account when each result has a high ambiguity in distance. The following discussions explain how these issues are managed in this study. 2.1
In the proposed method, The RSSI-ranging is performed using the RSSI observed
at the terminal. In order to consider the ambiguity in the estimated distance,
a evaluation function based on a Gaussian function is used, and the sum of
the evaluation function corresponding to all beacons is used to enhance the
localization accuracy. A conceptual sketch of the RSSI-ranging using a pattern
diversity is shown in Fig. 1. In this study, the multibeam feed network developed
in our previous work [
Distance-Based Localization Using Reliability Function
Even though the RSSI-ranging accuracy is improved by using the pattern
diversity technique, there still exists ambiguity in the estimated distance. The
RSSI-ranging method tends to incur a large error due to multipath-fading e ect
especially when the distance between the beacon and terminal is large. To
mitigate the e ect of the fading on the localization, the beacon with higher RSSI is
more trusted than other beacons, and this idea realizes further improvement in
the localization accuracy. The second key idea is a weighting scheme, which is
introduced to consider the reliability of the estimated distances from all beacons.
An evaluation function based on a Gaussian function is introduced to consider
the ambiguities in the estimated distance. In this scheme, the higher reliability
value is given for the closer beacon, while the lower value is given for the distant
beacon. The distance-based reliability function W is de ned as the reciprocal of
the power p of the estimated distance. p corresponds to the path-loss and this is
the simplest path-loss model commonly used in the various situations including
the indoor and outdoor environments [
Franging =
N X W (n)franging(n) n=1 (1) Where, N is the number of Tx and franging is evaluation function of RSSIranging. The summation operation is used to take all evaluation functions into account. The multiplication is not used because the correct solution disappears if one of the functions strays signi cantly. Finally, the receiver's location is estimated by searching the point, where the sum evaluation functions is maximum.
Tx2
1.6 m Obstacle
Beacon
Tx
Rx In this experiment, we performed indoor localization experiment using the BLE beacons and smartphone terminal, and evaluated the localization accuracy. To verify the performance of the ranging scheme without the e ect of the human body, we conducted the experiment with holding the terminal above the head of a person. The receiver's height is about 2 m. The experiment environment is detailed in Fig. 2. This experiment was conducted in a 25 m 37 m indoor environment. The measurements were performed at 36 points at 5 m interval in both vertical and horizontal directions. The metallic obstacles are intentionally located to consider the e ect of scatterers whose heights are about 1.6 m. The beacon transmitters were placed at the four corners of the measurement area. The locations of the transmitters are Tx1 (0.57 m, 0.55 m, 2.91 m), Tx2 (0.58 m, 24.5 m, 1.65 m), Tx3 (36.5 m, 24.4 m, 1.65 m), Tx4 (36.4 m, 6.55 m, 1.65 m). Besides, All transmitters direct the center of the measurement area. Fig. 3 shows a photo of the beacon transmitter and receiver used in the experiment. 0 0
Max Mean only only 5 Estimation error [m1]5 10 20 The receiver is commercially available smartphone without any alteration. As a transmitter, we used a device, which has a 3-element patch array antenna and transmits 4 beacon signals with 4 di erent patterns. 4-port feed network is connected to the transmitter via SP4T switch and the signal is sequentially switched all ports. The set of the beacon signal consists of 3 frequencies (advertising channels), i.e. 2.402 GHz, 2.426 GHz, and 2.480 GHz, and is transmitted by every 20 ms. 3.2
Fig. 4 shows the cumulative distribution function (CDF) of the localization error with the diversity scheme. We also show the CDFs of the localization errors when RSSI-ranging is performed using the average of all ports, only port h180 1, and only port h90 1 instead of the diversity scheme. The average and median localization errors of the mean-RSSI based ranging are 8.89 m and 7.99 m, respectively. On the other hand, the average and median localization errors of the diversity scheme are 7.77 m and 6.13 m, respectively. Therefore, it is found that the localization accuracy can be improved by selecting the port with the highest RSSI. Note that the diversity scheme described above is adopted in the following results. 3.3
Result 2: E ect of Distance-Based Reliability Function on Localization Accuracy The localization using the evaluation functions discussed above was evaluated, where the average of the Gaussian function is 0, and the variance is set to 100 m in this evaluation. In the proposed method, the constant p must be determined in advance of the localization. In this study, the two values, p = 0 and p = 0:8, were tested, where p = 0 means all estimated distances are equally trusted because weighting value is always 1 independently to the distance. Fig. 5 shows CDF of localization errors when p = 0:8 and p = 0. When the value is set to p = 0, the average and median errors are 7.77 m and 6.13 m, respectively. When the value is set to p = 0:8, the average and median errors are improved to 6.70 m and 4.62 m, respectively.
100 75 ] [F% 50 D C 25
With weight (p = 0.8)
Without weight (p = 0)
This paper has studied the accuracy improvement method of the RSSI-ranging based localization, where the pattern diversity and distance-based reliability function are jointly used. The experiment in the indoor environment revealed that the average and median errors are improved by 1.22 m and 1.86 m, respectively, compared to that of the non-diversity scheme. Moreover, the distancebased reliability function improves the average and median errors by 1.07 m and 1.51 m, respectively. Currently, the combination of the proposed method and DOD-based localization is experimentally studied for the further accuracy improvement, and a part of this result will be presented in the conference.