Sie sind auf Seite 1von 4

Passenger’s location estimation using Kalman Filter for Beacon fare

collection in a Wireless Low Floor Tram

Kiwoong Jung1, Pedro B. V. Bermudez2, HyeonCheol Hwang3, YongGuk Oh4, Jaeho Kwak5
Hanyang University 1, UST 2, Korea Railroad Research Institute 3, 4, 5
kiwoong464@krri.re.kr1, pedro@krri.re.kr2, hchwang@krri.re.kr3,
yongkuk5@krri.re.kr4, jkwak@krri.re.kr5

Abstract properly, every passenger has to tap at least once to


check-in and once more to check-out. It seems to be
BLE (Bluetooth Low Energy) is drawing attention satisfactory number of times, however in rush hours
as it is a new technology capable of communicating or when carrying luggage it can be troublesome.
using low energy and doesn’t require pairing as its Our idea is to propose an automatic fare collection
classic version. These two facts are important for IoT system able to pay fare ticket only using smartphone
applications making BLE a good candidate to be by its Bluetooth Low Energy function.
chosen over other communication technologies (Wi-
Fi, ZigBee, and Bluetooth Classic). One key point of 2. Bluetooth Low Energy (BLE)
the automatic fare collection system is the passenger
detection. As passenger nowadays are likely to carry Originally designed by Nokia as a low power
smartphones with embedded BLE function. Therefore, consumption communication protocol and named as
first in this paper will be explained how Bluetooth Wibree, Bluetooth Special Interest Group included
Low Energy can be used in public transportation this standard in Bluetooth 4.0 and changed its name to
systems such as in a Low Floor Wireless Tram to Bluetooth Low Energy [1]. The adoption rate was
check if someone took in fact a public transportation high making Bluetooth Low Energy embedded
vehicle (such as: bus, train, metro or tram) or if he or function present in almost all nowadays smartphones.
she is just nearby (waiting for the next vehicle for Comparing with Bluetooth Classic it doesn’t
example). Next, it will be explained how to improve require pairing process, making it a good candidate to
Bluetooth Low Energy passenger’s location detection be chosen in Internet of Things applications such as
when using RSSI (Receiver Signal Strength Indication) the one described in this paper.
for this task.
3. BLE Beacons
Keywords: Bluetooth Low Energy, Kalman Filter,
Automatic Fare Collection System, iBeacon, Receiver BLE Beacons are basically BLE transmitters
Signal Strength Indication which broadcasts packets to nearby devices using
2.4GHz waveband. This 2.4GHz waveband is divided
1. Introduction into 40 channels (each channel covers 2MHz interval)
and advertising occurs in three different channels: 37,
For a long period of time, public transportation 38 and 39 [2] as it can be seen:
fare payment relied on a manual method, a worker had
the only function of collecting ticket or money. As
times passed by, automatic fare collection systems
were invented in public transportations to improve
payment method. One advantage of it is the digital
payment, a transportation card with credits started
being used, and consequently, reduced required time
at stops. However, this method still have some
inconveniences.
First, it’s necessary to recharge transportation card
and, unfortunately, payments options are limited.
Figure 1: Wi-Fi and BLE sharing 2.4 GHz
When recharging sometimes only coins are accepted
waveband
or sometimes only cash, and so on. Second, the public
transportation demands tapping operations to work
Currently there are four important BLE Beacons
protocols: iBeacon, Eddystone-URL, Eddystone-UID
and AltBeacon. Each one of these protocols sends data
differently. In the iBeacon protocol, which one is used
in this project, advertised data packet frame [2] as
shown in Figure 2:

Figure 3: Fare collection system model

As explained in the “BLE Beacons” section of this


paper, in iBeacon protocol one of the transmitted data
is the “TX power”. It indicates the RSSI (Received
Figure 2: Advertising mode data frame Signal Strength Indication).
By analyzing RSSI we can estimate someone’s
Advertising data packet is 31 bytes long and smartphone proximity with the BLE beacon. The
composed of 5 smaller fields. The first field is higher RSSI value is, closer receiver is to advertising
“iBeacon prefix”, this field is 9 bytes long used to mode device, and in this case, more likely is to
specify beacon proprieties, it includes “advertising someone be inside the vehicle. On the other hand,
flags”, “advertising header”, “company ID”, “iBeacon when beacons stops being detected by passenger’s
Type” and “iBeacon Length”. Following there is smartphone it is likely that passenger arrived at
“Proximity UUID” field, this field is 16 bytes long destination and took off the vehicle. However only
and has BLE beacon device unique ID. Next, “Major” estimating by RSSI is not enough to proceed with
and “Minor” fields, each one of them, 2 bytes long. payment automatically so is necessary to have a
“Proximity UUID”, “Major”, “Minor” roles can be confirmation mechanism.
seen in table [3]: This way, when passenger’s smartphone detects
vehicle’s beacon device and RSSI is sufficiently high,
Table 1: iBeacon parameters description higher than stipulated threshold, a pop-up message
will appear on the passenger smartphone screen
asking if in fact he or she boarded a vehicle. After
passenger confirmation, smartphone keeps scanning
periodically during the trip to check if the beacon still
nearby. Eventually when passenger take off, beacon
will stop being detected, so a new pop-up message
will appear on smartphone screen to confirm
Lastly in the data packet frame there is “TX power” passenger arrival on destination. Finally, trip itinerary
field which indicates RSSI (Received Signal Strength will be sent to billing server and fare calculated price
Indication), it can be used to estimate distance sent back to smartphone to start payment procedure.
between a BLE device in advertising mode and BLE Previous algorithm is described in below figure:
receiver.

4. BLE Beacon Fare Collection System


One key point of the automatic fare collection
system is the passenger detection. As passengers
nowadays are likely to carry smartphones with BLE
function and BLE beacons were attached to the
vehicle it is possible to check when a person took and
took off the vehicle. The system basic configuration is
the following one:

Figure 4: Automatic fare collection system


passenger detection and payment procedure
𝑅𝑆𝑆𝐼(𝑑0 )−𝑅𝑆𝑆𝐼(𝑑)
5. Wireless Low Floor Tram d = 10 10𝛾 (2)

The Wireless Low Floor Tram is a tram designed In this case, more important than distance
by Korea Railroad Research Institute. It was calculation correctness, we want to have a correct IN
developed to be completely sourced by a 196 kWh Li- or OUT estimation. This way, a test composed of two
polymer battery and achieve a top speed of 70 km/h. steps were set. The first step of the test was to measure
The main features are its ability to run 35 km distance RSSI value while smartphone was inside of the
with a single charge, complete curves as sharp as 25 vehicle, and the second step was to measure RSSI
m radius and, inside the tram, the indoor noise level is value when smartphone was outside the vehicle (1 m
below 70dB. Because of its low floor property it away). Tests elements were: tripod, smartphone,
provides accessibility for wheelchair users and elderly beacon and Wireless Low Floor Tram. Test
people. The tram consists of 5 cars and in each end of configuration pictures can be seen below:
the tram there is a driving cabin, so it’s a bi-directional
vehicle.

Figure 6: IN and OUT situations, yellow circles


indicate smartphone; blue circle, beacon
Figure 5: Wireless Low Floor Tram developed by
KRRI being tested at Osong track To measure RSSI, we monitored RSSI through the
time using manufacturer recommended application.
Even though there are 388 cities in the world that Sampling rate was 1 every 5 seconds, total
has trams, unfortunately in Korea there isn’t any city. measurement time was one minute, obtained results
This scenario is about to change as there is the can be seen below:
possibility of Daejeon implementing Korea’s first
tram in its metropolitan Line Number 2. Although
paper results were obtained for this tram specifically,
similar approach and ideas of this project can be used
in other transportation systems, such as: trains, metro,
buses.

6. Method
As RSSI is used in “BLE Beacon Fare Collection
System”, it is important to know that, theoretically,
RSSI is attenuated following log-distance path loss
model:

𝑑 Figure 7: RSSI - IN and OUT situation


PL = P𝑇𝑥𝑑𝐵𝑚 − 𝑃𝑅𝑥𝑑𝐵𝑚 = 𝑃𝐿0 + 10𝛾 log10 ( ) + 𝑋𝑔
𝑑0
(1)
RSSI values are instable, this is due to the fact
Path loss (in equation 1 nominated as “PL”) is the BLE transmitted power is limited and noise is
attenuation of an electromagnetic wave when it is inevitable as multipath destructive interference is
propagated through the space. For a certain distance present. In fact, Inside the Wireless Low Floor Tram,
“d”, path loss value can be obtained comparing it with RSSI values oscillated more frequently than outside
a reference value (“PLo”). Furthermore, γ is the path ( σin = 6.14 and σout = 1.20 ), by “ σ ” we mean
loss exponent and depends with the transmission standard deviation.
medium. “Xg” is a Gaussian random variable that
represents the attenuation caused by fading [4]. If we 5. Results
want to calculate the distance by RSSI in a “no fading”
scenario, is possible to use the following equation: Therefore, to improve our measurements results,
Kalman filter was used. This filter was developed by
R.E. Kalman as a regression filter that tracks the 6. Conclusion
condition of linear and non-linear states that contain
noise. The filter predict and revise states of value past, First, in this paper it was introduced how
present and future. It is useful to predict and track automatic fare collection system works and its
about model system [6]. importance. Next, it was explained how Bluetooth
The prediction phase is expected when user enters Low Energy technology works (including BLE
the input for anticipated state in the previous time beacons) and how they can used with automatic fare
period. The compensation phase is stage for collection system. Later, it was explained how to
estimating the exact state of the process based on detect IN and OUT situations and improved results
processed estimation state and real state. The were obtained by applying Kalman Filter.
estimated process states according to time are Even though classification still simple as a linear
expressed as average and distribution [7]. classification is applied, in future works we expect to
use more sophisticated algorithms, such as neural
network. This project is interesting as it concerns two
nowadays IT important areas: Internet of Things and
Machine Learning. As it’s likely to have in the future
an automatic fare collection systems when using
public transportation is important to think about its
development, as it is a profitable market.

This research was supported by “Demonstration of


Figure 8: Kalman filter algorithm
Wireless Low Floor Tram” Project (17RTRP-
13137310-01) funded by Korean Ministry of Land,
Analyzing filtered signals is easier to classify
Infrastructure and Transport.
correctly if someone is IN or OUT situation as in this
case RSSI values for these two situations are more
spaced one to each other. In the classification process, References
if measured RSSI value is higher than a threshold, the
place where RSSI was measured is considered to be [1] D. Robert, T. Kevin, W. Chris, C. Carles. “Gettin
inside the tram, below threshold, outside the tram. g Started with Bluetooth Low Energy”. O’Reilly Medi
a. April 2015.
Threshold can be estimated as a value equidistant of
the curves for the two situations, as it follows: [2] Bluetooth Core Specification v 4.2, Vol 2, Part E,
chapter 7.8.5

[3] Apple. Getting started with iBeacon: https://devel


oper.apple.com/ibeacon/Getting-Started-with-iBeacon.pdf
(Accessed 2017-08-30)

[4] Sampei, S. Applications of Digital Wireless Techn


ologies to Global Wireless Communications, 2000. Feh
er/Prentice Hall. Wireless and Digital Communication
Series.

[5] Z. Paul; M. Howard (2000). Fundamentals of Kal


man Filtering: A Practical Approach. American Institut
e of Aeronautics and Astronautics, Incorporated. ISBN
978-1-56347-455-2.

Figure 9: IN/OUT classification by RSSI, after [6] G. Welch and G. Bishop, “An Introduction to the
applying Kalman Filter Kalman Filter,” UNC-Chapel Hill TR 95-041, July 20
06.
In this case, after the signal is filtered for 30
[7] Chang-Pyo Yoon and Chi-Gon Hwang, 2015, "Efficient
seconds, the signal becomes stable, signal standard
∗ indoor positioning systems for indoor location-based
deviation values get much lower after it (σin = 0.61 service provider," Journal of the Korea Institute of

and σout =0.10). Comparing to previous situation Information and Communication Engineering, Vol. 19, No.
where from 30 seconds to 60 seconds the measured 6, pp. 1368~1373.
signals were still unstable (σin = 5.40 and σout =
1.15) we can notice how effective Kalman filter was
in filtering RSSI signal.

Das könnte Ihnen auch gefallen