See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/340250091

Smart Parking Sensors: State of the Art and Performance Evaluation

Article  in  Journal of Cleaner Production · March 2020

DOI: 10.1016/j.jclepro.2020.121181

CITATIONS READS

40 5,352

5 authors, including:

Toni Perković Petar Šolić

University of Split University of Split
70 PUBLICATIONS   368 CITATIONS    93 PUBLICATIONS   1,545 CITATIONS   

SEE PROFILE SEE PROFILE

Hamid Zargariasl Joel Rodrigues

Ericsson Nikolca Tesla Senac Faculty of Ceará
19 PUBLICATIONS   69 CITATIONS    1,071 PUBLICATIONS   28,945 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Precision local positioning mechanism in underground mining using IoT-enabled WiFi platform View project

Internet of Things (IoT) and its applications: Architectures, Communication protocols, Resource management View project

All content following this page was uploaded by Toni Perković on 18 April 2020.

The user has requested enhancement of the downloaded file.

 Smart Parking Sensors: State of the Art and Performance Evaluation

T. Perkovića,∗, P. Šolićb , H. Zargariaslb , D. Čokob , Joel J. P. C. Rodriguesc,d

a University Department of Forensic Sciences, University of Split, Croatia
b Department of Electrical Engineering, FESB, University of Split, Croatia
c Federal University of Piauı́ (UFPI), Teresina - PI, Brazil
d Instituto de Telecomunicações, Portugal

Abstract
Smart parking systems present one of the essential infrastructure concepts that enable Internet of Things (IoT) in Smart
cities. Finding a free parking lot plays a role in reducing traffic congestion, gas emissions, and increasing the quality
of life of people living and working in it. Therein, the fundamental part of every smart parking system is the detection
of vehicle presence, which is usually employed by devices comprised of power-hungry sensors. This paper gives an
extension to state-of-the-art by a systematic in-depth overview of technologies used for the smart parking detection
realization consuming mW of power. Deeper insights on the real-scenario performances and power consumption of
most popular sensor devices and Low Power Wide Area (LPWA) radio technologies available today (LoRa, Sigfox
and NB-IoT) are provided. The results show that based on the architectures of IoT system, lowest consumption is
for LoRa devices. Further, analysis of power consumption of commercial LPWA-based Smart parking sensor device
is provided along with battery estimation lifetime, which is especially important for the deployment of future smart
parking solutions. Battery lifetime heavily depends on the number of parking lots exchanges and based on the less
frequent changes, the estimated battery lifetime is approximately 7 years. Inspired by the limitations of power-hungry
and relatively expensive smart parking sensor devices, two strategies for the optimization are proposed: first one is
based on the premise where a drop in received signal strength of the LPWA device can serve as the presence of the
vehicle in the parking lot, while the second one proposes a big picture on a novel architecture for harvesting the
surrounding energy and using the same for circuitry wake-up therefore saving the energy.
Keywords: Sensor performance, Smart parking, LPWA, IoT power consumption

1. Introduction

One of the most important topics addressed by the European Commission and most nations in the world is the
development of an urban city model aimed at increasing the quality of life of people working and living in them.
Smart and Sustainable Mobility is one of the central concepts in the vision of the Smart City, where Internet of Things
plays very important role (Perković et al., 2019; Nizetic et al., 2019).
Fundamental problem underlying the Smart City concept is the reduction of traffic congestion in the city. Con-
sidering the fact that on average drivers spend about 7.8 minutes finding a free parking space, major contributor to
reducing traffic congestion would be to develop smart parking systems aimed at finding a free parking space more
efficiently. Such problems are even more emphasized by knowing that approximately 30% of daily traffic conges-
tion is caused by drivers looking for a free parking space (Geng and Cassandras, 2013). All this further increases
the drivers’ frustration with accomplishing daily activities, but moreover, it dramatically increases fuel consumption
which contributes to the air pollution in crowded cities.
Recent advances of wireless sensor networks, along with low-cost sensor devices, Web technology and pervasive
computing gained momentum to the expansion of IoT ecosystem (Bibri, 2018; Cruz et al., 2019). Combined with

∗ Correspondingauthor
Email address: [email protected] (T. Perković)

Preprint submitted to Journal of Cleaner Production April 18, 2020

 Network server
Network server / cloud sends messages
for visualization

occupied
parking lot

free
Gateway parking lot

Figure 1: Smart parking sensor devices communicate with network server / cloud using a gateway (multiple gateways can be used). From there,
network server forwards incoming data for visualization.

big data analytics, IoT started to stimulate the advancement of solutions that enhance and/or complement car traffic
monitoring, especially in overpopulated cities aimed at increasing smart sustainable mobility (Zanella et al., 2014). In
recent years, a number of solutions have been proposed aimed at resolving the problem of finding free parking spaces
aimed at establishing a green environment and increasing the quality of life in Smart cities (Lin et al., 2017; Kotb
et al., 2016).
In essence, the majority of these solutions are based on a mobile application through which users receive infor-
mation about free parking spaces (Figure 1). These solutions range from Parking Guidance and Information (PGI)
to even Parking Reservations (PRS) Systems delivering information to the drivers over the Internet about available
parking space (Kotb et al., 2017). By taking into account the distance and the total number of free places in a parking
lot a mobile application will automatically find a free parking lot using metrics at the least price (Pham et al., 2015).
Within the context of IoT, the implementation of solutions for the detection of free parking lots is based on the
implementation of devices that can sense the environment (in this case the presence of vehicles) and send this data to
a central server for further processing (e.g. via radio channel). Selection of the appropriate sensor device for vehicle
detection rather depends on the requirements of parking lot, focusing at solutions that preserve high accuracy while
reducing the overall cost (Al-Turjman and Malekloo, 2019). Depending on the existence of external power source,
these sensor devices can be classified into active and passive ones. These devices (sensors) range from light sensors,
distance sensors based on either infrared or ultrasound, magnetometers, and even combinations of different sensor
devices (Šolić et al., 2019; Lin et al., 2017). Passive infrared sensors work as heat detectors from surrounding objects,
while the vehicle passes in the vicinity of the sensor device (Kotb et al., 2017). Active infrared sensor device estimate
the distance between the sensor device and a surrounding object in front of them (Lin et al., 2017). On the other hand,
ultrasound sensors measure round-trip time of ultrasound reflected from an object positioned in front of them (Al-
Turjman, 2018). Cameras with video analysis can also be employed to detect the presence of vehicles placed in a
parking lots (Wahyono and Jo, 2017). Magnetometers detect the change in magnetic field once the vehicle (metallic
object) passes above them (Arab and Nadeem, 2017). Light Dependent Resistor detect luminous intensity change,
while piezoelectric sensors detect vehicle passing over from the pressure/vibration of the vehicle (Al-Turjman, 2018).
Although sensor device manufacturers guarantee several years of lifetime for mostly battery-powered devices, the
consumption of radio front-end, the sensors themselves, and microcontrollers make them extremely power hungry.
However, compared to their functionalities such sensors are rather expensive.
Nevertheless, from the perspective of end user, parking space status, i.e. occupied or available, can be coded
with only one bit of information. On such basis, this paper describes an overview of sensing technologies used for

2
detecting the occupancy of parking lots focusing on performances of most popular sensors used nowadays as well as
their power consumption analysis. In particular, to extend existing contributions, this paper provides an analysis on
the performance of the photodiode, Light-dependent resistor (LDR), Infrared LED, Ultrasonic, LiDAR and magnetic
sensors. The data is delivered in terms of vehicle presence, accuracy and power consumption that can be easily
transferred as a feasibility information to integrate in smart parking sensors. Moreover, performances of a LPWA
radio communication technology, such as such, as LoRa, NB-IoT and Sigfox (Centenaro et al., 2016; Raza et al.,
2017), used in smart parking detection systems today are provided in terms of power consumption and availability.
As a case study, performances of a commercial LoRaWAN Smart parking system are provided to better understand
power requirements/protocols of smart parking devices used today. To cope with the power consumption issues, two
strategies for the optimization are proposed. First one shows that information on the received signal strength of the
LPWA device could serve as an indicator of the occupancy, where a drop in signal strength can be detected on the
receiving side when the vehicle is located above the sensor device (Solic et al., 2019). The premise is based on a
simple but effective principle: being placed in the parking lot, the device is provided with a radio front-end that is
detuned by the presence of the electromagnetically harsh environment, i.e. vehicle. Second one proposes a novel
device that compensates expensive, mainly battery-operated and power hungry LPWA-based smart parking sensors.
In particular, a big picture on novel architecture for harvesting the surrounding energy and using the same for circuitry
wake-up therefore saving the energy is proposed.
The objective of this work is to analyze off-the-shelf sensor devices that can be used for building the smart parking
node devices in terms of their detection accuracy and power consumption. In addition, the power requirements of
commercial and most prominent LPWA technologies are also investigated, such as LoRa, Sigfox and NB-IoT, that
can be used as a radio interface for building the same smart parking devices. In depth analysis of commercial LPWA
smart parking detector in terms of consumption and lifetime duration is also provided. The related parameters affecting
the power consumption resulted in two potential strategies that may extend battery lifetime of smart parking sensor
device.

2. Sensing the Occupancy - Smart Parking Sensor Performance Evaluation

Smart parking systems hold their place as one of the essential infrastructures that enable Internet of Things in
Smart cities. Hence, the fundamental part of the system lies in the vehicle detection placed in the parking lot. In
this section a range of sensing technologies is introduced for the detection of parking lots. More particularly, focus is
placed on the most popular sensor devices used in real parking scenario, such as a photodiode, ultrasound, infrared,
magnetometer, and Time-of-Flight sensors. At the end, in depth analysis of power consumption of sensor devices is
provided along with battery estimation lifetime, which is especially important for the deployment of smart parking
sensor devices.

2.1. Smart parking sensor detection

For the purpose of evaluating the performance of sensor devices used in smart parking environments, including
the power consumption, we utilized the following sensors: magnetometer HMC5883L 1 , IR sensor, Time-of-flight
WAVGAT VL53L0X, Light Dependent Resistor, Photodiode and Ultrasound HC-SR04 which were connected to the
Arduino Nano board (Figure 2) operating at 5V. First, sensor devices were placed in several open/free parking lots to
gather sensor information about a free parking lot. After that, sensor devices were placed on the ground below several
vehicles with various distance chassis to obtain sensor information status.
As can be seen in Figure 3, 1023 presents the max. value that the analog-to-digital converter can give on the output
(10-bit ADC converter) which indicates that the photodiode is in saturation. When photodiode enters saturation, it
can be assumed as an open parking space. However, some measurements indicate an open parking space (value 1023)
although the vehicle is placed above the photodiode. It is assumed that the reflections of the light from the surrounding
vehicles increase the light level received by the photodiode.
During the calibration procedure, ultrasound detector showed values around 140, as can be seen in Figure 4(left).
Unfortunately, during the experiments in realistic scenario, some measurement showed false predictions (smaller than
the threshold value). This indicates that the used ultrasound sensor can not be reliable in every vehicle detection.

1 https://www.farnell.com/datasheets/1683374.pdf

3
 Figure 2: Smart parking sensor devices connected to Arduino Nano board.

1050

no vehicle present
1023
Photodiode sensor output value

1000

950

900

850
10 15 20 25 30
distance (different vehicle chasis to the ground) [cm]

Figure 3: Sensing vehicle presence by Photo-diode.

350 350
Ultrasound sensor output value

300
LIDAR sensor output value

300

250
250

200

200
150
no vehicle present
150
100

50 100
10 15 20 25 30 10 15 20 25 30
distance (different vehicle chasis to the ground) [cm] distance (different vehicle chasis to the ground) [cm]

Figure 4: Sensing vehicle presence by (left) Ultrasound detector and (right) LiDAR. The read message via LiDAR is ”Out of Range” in any open
lot.

4
 -1000 0
no vehicle present
-500

-1500 -1000
-1093 no vehicle present
z-axis output data

z-axis output data

-1500

-2000 -2000

-2500

-2500 -3000

-3500

-3000 -4000
15 20 25 30 35 40 10 15 20 25 30
distance (different vehicle chasis to the ground) [cm]
distance (different vehicle chasis to ground) [cm]

Figure 5: Output of magnetometers sensor devices placed at the parking lot.

120 50

100 40
IR sensor output value

IR sensor output value

80
30

60
20
40
10
20

0
0
no vehicle present
no vehicle present
-20 -10
15 20 25 30 35 40 10 15 20 25 30
distance (different vehicle chasis to ground) [cm] distance (different vehicle chasis to the ground) [cm]

Figure 6: Output of IR proximity detector placed at the parking lot.

Figure 4(right) states that the LiDAR, as a sensor, can be considered as a good candidate so that the absence of a
vehicle is explicitly shown. There is not any reference line because the sensor issues a message of ”Out of range” for
empty parking lots.
In some scenarios it can be seen that measurement output from a sensor device indicates free parking space,
although the sensor is placed under the vehicle. Particularly, the measurement results from the magnetometer and
Infra-red sensor in some measurements indicate a non-occupied parking lot (Figures 5 and 6). These inaccurate
results largely depend on the distance between the chassis and the ground, along with metallic structure of the chassis.
As can be seen, if sensor devices are placed from the sensor device, a larger likelihood of inaccurate parking lot
occupancy will be obtained.
In next section power consumption requirements are provided which define research directions that may lead to
the further optimization of the sensing devices.

2.2. Power Consumption Analysis

In order to compare the power consumption of these sensors, sensor devices were connected to an Arduino Nano
device running a different sketch for each individual sensor. The power consumption was derived from average voltage
and current supplying the Arduino Nano at the +5 V pin. An LED was connected to digital output D3 as an indication
of an object. A different threshold value was set for each individual scenario.
5
 Figure 7: Laboratory setup comprised of sensor devices with microcontroller and metallic plate.

Photodiode (visible spectrum). The photodiode was connected from the +5 V pin through a 10 kΩ pull-down
resistor to GND. The voltage drop on the resistor was measured at analog input A1. A threshold for indication of
an object was set to 900 ADC units, turning the LED on when the photodiode was completely shaded. The average
measured current consumption was 13.47 mA at a supply voltage value of 5.52 V, which yields an average power
consumption of 74.35 mW.

Light-dependent resistor (LDR). The LDR was connected from the +5 V pin through a 1 kΩ pull-down resistor
to GND. The voltage drop on the resistor was measured at analog input A0. A threshold for indication of an object
was set to 650 ADC units, turning the LED on when the LDR was completely shaded. The average measured current
consumption was 18.85 mA at a supply voltage value of 5.47 V, which yields an average power consumption of
103.11 mW.

Infrared LED and photodiode. The photodiode was connected from the +5 V pin through a 330 Ω pull-down
resistor to GND. The voltage drop on the resistor was measured at analog input A3. An infrared LED was connected
from digital pin D7 through a 330 Ω pull-down resistor to GND. The ambient infrared light was measured, stored and
subtracted from the result to increase the accuracy of the system. In this manner only the infrared light reflected from
the obstacle was taken into account. A threshold for indication of an object was set to 30 ADC units. The average
measured current consumption was 19.42 mA at a supply voltage value of 5.46 V, which yields an average power
consumption of 106.03 mW.

Ultrasonic sensor. The HC-SR04 ultrasonic sensor was connected to the Arduino Nano (trigger to pin D2, echo
to pin D1). The emitted ultrasonic pulse length was set to 10 µs. A threshold for indication of an object was set to
100, corresponding to approx. 10 cm distance. The average measured current consumption was 16.25 mA at a supply
voltage value of 5.49 V, which yields an average power consumption of 89.21 mW.

LiDAR sensor. The Vl53L0X micro smart LiDAR sensor was connected to the Arduino Nano via I2C bus. A
threshold for indication of an object was set to 200, corresponding to approx. 10 cm distance. The average measured
current consumption was 27.31 mA at a supply voltage value of 5.29 V, which yields an average power consumption
of 144.47 mW.

3-axis magnetic sensor. The QMC5883 3-axis magnetic sensor was connected to the Arduino Nano via I2C
bus. Only the z-axis value was read. A threshold for indication of an object was set to 2800. The average measured
current consumption was 14.98 mA at a supply voltage value of 5.51 V, which yields an average power consumption
of 82.54 mW.
The measurement results show that the minimal power consumption can be obtained when using a photodiode (in

6
 Table 1: Comparison of different LPWA development boards

MKR WAN 1300 (LoRa) MKR FOX 1200 (Sigfox) MKR NB 1500 (NB-IoT)

Manufacturer Arduino Arduino Arduino

Core SAMD21 ARM Cortex M0+ SAMD21 ARM Cortex M0+ SAMD21 ARM Cortex M0+
Clock Speed 32.768kHz (RTC), 48MHz 32.768kHz (RTC), 48MHz 32.768kHz (RTC), 48MHz
Board Power Supply 3.3 V 3.3 V 3.3 V
Working region EU EU (RCZ1) EU
Frequency 863-870MHz Uplink center frequency B3 (1800 MHz), B8 (900 MHz)
868.130MHz, Downlink center and B20 (800 MHz)
frequency 869.525MHz
Radio modem Murata CMWX1ZZABZ Atmel ATA8520 U-Blox SARA-R410M-02B
Antenna power 2dB 2dB 2dB
Current Consumption (in TX) 128 mA 45 mA 20-60 mA (max. 190 mA)
TX period 3 sec 15 sec 5-10 sec

the visible spectrum). This is due to the operating principle of the photodiode, as it is reverse biased so the current
through the diode is very small, almost negligible. This mode of operation (often called photoconductive as opposed to
photovoltaic in forward-bias) also causes the voltage drop on the pull-down resistor to have almost linear dependence
on the illumination of the diode. Unfortunately, the vehicle occupancy measurement results indicate that photodiode,
although gives the best results in terms of consumption, does not output good quality vehicle presence results. The
second best sensor in terms of power consumption is the 3-axis magnetic sensor, followed by the ultrasonic sensor.
The other sensors that were tested exhibited a power consumption above 100 mW for 5V input.

3. Power Requirements of Commercial LPWA Technologies

In this section comparison between three LPWA technologies is given: Sigfox, NB-IoT and LoRaWAN. Three
boards from the same manufacturer were selected. Specification of every LPWA node is given below, which is
followed by measurements of power consumption. The results of measurements will give us a direction into selecting
the most appropriate LPWA technology suitable for our needs, taking into account power consumption.
In this section comparison between three LPWA technologies is given: Sigfox, NB-IoT and LoRaWAN. Three
boards from the same manufacturer were selected. Specification of every LPWA node is given below, which is
followed by measurements of power consumption. The results of the measurements will give us a direction in selecting
the most appropriate LPWA technology suitable for our needs, taking into account power consumption.
Table 1 gives a comparison of MKR-based LPWA boards that were used for test setup: MKR WAN 1300 (LoRa
Connectivity), MKR FOX 1200 (Sigfox) and MKR NB 1500 (NB-IoT). All three boards are based on Atmel SAMD21
Cortex-M0 32-bit low power ARM MCU. MKR WAN 1300 uses Murata CMWX1ZZABZ2 LoRa module powered
with operating Voltage 3.3V. MKR FOX 1200 uses Atmel ATA85203 single-chip Sigfox RF transmitter with circuit
operating voltage 3.3V, while MKR NB 1500 with U-Blox SARA-R410M-02B4 modem and operating Voltage 3.3V.
Figure 8 shows test setup used for our measurement. Every board has Micro U.FL connector for antenna connec-
tion. In our case we used the antenna with 5dbi Gain and SMA Male Connector. To measure current consumption,
every board was connected to Current Ranger, and powered from external power supply of 3.4V. An oscilloscope
was connected to Current Ranger to capture detailed measurements of current consumption. In addition, the Anritsu

2 https://wireless.murata.com/pub/RFM/data/type_abz.pdf
3 http://ww1.microchip.com/downloads/en/devicedoc/atmel-9372-smart-rf-ata8520_datasheet.pdf
4 https://www.u-blox.com/sites/default/files/SARA-R4_DataSheet_%28UBX-16024152%29.pdf

7
Figure 8: LPWA development boards connected to external power supply and Current Ranger to measure consumption in active and inactive mode:
(left) MKR WAN 1300 - LoRa module, (center) MKR FOX 1200 - Sigfox (right) MKR NB 1500 - NB-IoT.

50
30 60

40
Consumption (mA)

Consumption (mA)
Consumption (mA)

45
20
30

30
20

10
15
10

0 0 0
1 2 3 4 5 10 20 30 40 50 2 4 6 8
Time (s) Time (s) Time (s)

Figure 9: Consumption of LPWA development boards based on (left) LoRaWAN (center) Sigfox and (right) NB-IoT technology.

Figure 10: Output power of Anritsu MS2711E spectrum analyzer for (left) LoRaWAN (center) Sigfox radio and (right) NB-IoT technology.

MS2711E spectrum analyzer was utilized to capture channel power, central frequency for every node and occupied
bandwidth. In our code ArduinoLowPower library was utilized to get our board periodically into deep sleep.

3.1. Measurement results

Figure 9 shows the power consumption for LoRaWAN, Sigfox and NB-IoT development boards that were cap-
tured on the oscilloscope. To measure consumption of every device, output of Current Raider was connected to the
oscilloscope. The output of the signal was saved in a CSV file and further filtered in Matlab using a moving average
filter. As can be seen, all development boards have significant consumption in sleep mode (e.g. more than 1.6mA
for LoRaWAN module). Such a high consumption is due to soldering problem, and many forums discuss how to
decrease consumption by inserting hardware changes into the board, or in some scenarios, modules such as low power

8
TPL5110 is suggested that completely cuts off the consumption to the board.5 Consumption from LoRaWAN module
during transmission is around 35 mA, with 1.12 dBm output power and central frequency 850 MHz (Figure 10(left)).
Both transmission and reception time period last no longer than 3 seconds. It can be noted that both TX and RX period
depend on the spreading factor and bandwidth window. In the scenario shown in Figure 9, for choosing parameters
SF9BW125 airtime was around 206 ms6 . In active mode, it may be noted that Sigfox module transmits message three
times, which is characterized by Sigfox specification, and consumption is larger than 45mA during the transmission
period. Besides sending three messages it seems that prior going to sleep the device spends a large portion of time
waiting for a message from the base station which significantly increases the overall power consumption. Even if
Sigfox device does not check for downlink messages, active period during TX will be around 15 seconds. Anritsu
MS2711E spectrum analyzer measures the output power around 12.8 dBm with central frequency 868 MHz (Fig-
ure 10(center)). At the end, NB-IoT consumption shows to be very large during active period, larger than 20mA on
average for the period of 5-10 seconds, with spikes going over 65 mA during transmission. Anritsu MS2711E spec-
trum analyzer measures the output power around 15 dBm with central frequency 848 MHz (Figure 10(right)). Note
that in the sleep period microcontroller is in power-down mode, while radio is constantly powered and not switched
off. Regardless, in active mode consumption is larger than both LoRaWAN and Sigfox technology. In conclusion,
LoRaWAN currently presents as the best LPWA technology since it is infrastructure independent (both for end device
and gateway device), can be ”easily” installed, and power consumption is by far smallest compared to other LPWA
technologies.

4. Analysis of Commercial LPWA Parking Detection Systems

Libelium Smart Parking Sensor LoRaWAN Network and

base station Application server

Figure 11: Network architecture of Libelium Smart Parking sensors.

In this section details for connecting Libelium LoRaWAN-based Smart parking sensor devices are provided. Al-
though Libelium suggests Loirot as a LoRaWAN provider, good experience as well as documentation provided by
The Things Network (TTN) LoRaWAN system motivated us to pursue in establishing a connection of Smart Parking
sensor devices to their cloud system. Moreover, TTN allows us to forward all messages from Libelium smart parking
sensor to our personal server comprising Node-RED, InfluxDB and Grafana services for visualization and further
processing. Figure 11 shows the architecture of our Libelium Smart Parking system.
First step is to obtain LoRaWAN configuration parameters, i.e. parameters for LoRaWAN OTAA (Over the Air
Authentication) protocol - Device EUI, Application EUI and App Key. To accomplish this, it is necessary to connect
Libelium Smart Parking sensor device to the PC via USB cable and switch the node in the Boot mode (Figure 12).
Smart Parking devices already come with preloaded parameters, and they are essential since they have to be identical
to the parameters configured in The Things Network cloud application for the given sensor device. To accomplish
this, Libelium offers Smart Devices App written in the Java language for programming of their parking sensor devices.
After obtaining OTAA parameters from the device using Libelium Smart Devices App, in TTN console
(https://console.thethingsnetwork.org) user creates an application and configures the parameters to match OTAA pa-
rameters in Libelium Smart Devices App. After that, user disconnects Smart Parking sensor from the computer,
switches to App mode and pushes Reset button to start with the OTAA protocol with The Things Network base
station.

5 TPL5110 datasheet: http://www.ti.com/lit/ds/symlink/tpl5110.pdf

6 https://avbentem.github.io/lorawan-airtime-ui/ttn/eu868/13

9
 Figure 12: Libelium Smart Parking sensor device.

Table 2: (left) Libelium Smart Parking sensor packet header, (right) Uplink frames types.

Byte Bit Field Frame type num Description

0 7 Parking lot status Start frame 1 4 First frame sent by the node when starting with params settings
0 6 Battery state Start frame 2 5 Second frame sent by the node when starting with param settings
0 5 Configuration uplink acknowledgment Info frame 0 Inform about parkgin status change
0 4 Sensor recalibration Keep-alive frame 1 Device working since last reported status
0 3-0 Frame type Configuration uplink 2 Confirmation whether configuration downlink was applied
1 7-0 Sequence number RTC update request 7 Request for RTC sync once every day

Decoding the messages. Libelium has a very detailed description about payload coding in their documentation,
which allowed us to create our own decoder. As can be seen in Table 2, Upload message payload comprises 11 bytes,
out of which first 2 bytes correspond to the header byte, while 9 correspond to the payload itself. For example, bit 7
of 0 header byte corresponds to the parking lot status, while bits 0-3 of header byte 0 correspond to the Frame Type
(such as start frame, info frame or keep-alive frame - Table 2). Next, we give an example of a payload format decoder
created in The Things Network console for our application.
1 function Decoder(bytes, port) {
2 var bitIndex = 7;
3 var bitMask = 1 << bitIndex;
4 var result = bytes[0] & bitMask;
5 var parkingStatus = result >>> bitIndex;
6 var bitMaskFrType = 7;
7 var resultFrType = bytes[0] & bitMaskFrType;
8 var frameType = resultFrType;
9
10 return {
11 parkingStatus: parkingStatus,
12 frameType: frameType
13 };
14 }
Listing 1: Payload Decoder Function

A flow was created in Node-RED on our server side where the first block establishes connection to TTN, specif-
ically, to the previously created application in TTN cloud that receives messages from the Libelium Smart Parking
device. After that, all received messages are parsed and prepared for the InfluxDB. Next, Grafana was used to connect
to the InfluxDB for displaying received data as they appear in the database. Figure 13(left) shows parking lot status
for 5 sensor devices while Figure 13(right) shows RSSI of packets received during the period of 12 hours from one
device. As can be seen, when a vehicle is placed above the sensor device, a drop in RSSI is detected on the receiving
side, which could serve as an indicator of the occupancy.

10
 Free Occupied
Occupied
-70
Gateway 1
Parking lot 1 Gateway 2
Parking lot 2 -80 Gateway 3
Parking lot 3 Gateway 4
Parking lot 4
Parking lot 5
-90
Status

RSSI
-100

-110

-120
Free
-130
2:00 4:00 6:00 8:00 10:00 12:00 14:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00
Time (h) Time (h)

Figure 13: (left) Parking status of 5 Libelium parking sensor devices (right) RSSI captured on four LoRaWAN gateways from one Libelium Parking
sensor device.

140

250
120

200
MCU periodically wakes
up magnetometer
Parking lot free 100
Consumption (mA)

OTAA Start fr. 1 Start fr. 1 Info fr. Info frame

joint 60 seconds
Consumption (mA)

Start fr. 1 Start fr. 2 Parking lot occupied

Info frame
80
150

60

100
40

50 20

0
0
50 100 150 200 250 300 0 20 40 60 80 100 120 140
Time(s) Time(s)

Figure 14: (left) Libelium Smart Parking sensor consumption during the initial OTAA connection to the gateway. (right) Libelium Smart Parking
sensor consumption between two microcontroller wakeup periods.

4.1. Consumption and lifetime estimation

To estimate the lifetime of sensor device Libelium smart parking sensor was connected to the Current Ranger,
whereas Current Ranger was connected to the Oscilloscope to capture detailed measurements of current consumption.
Figure 14(left) shows consumption during the OTAA authentication procedure captured by the oscilloscope. Similarly,
as explained in Libelium documentation (Table 2(right)), after OTAA authentication, Smart Parking device sends Start
frame 1, which is followed by Start frame 2. After that, Info frame is sent. Next, sensor device enters sleep where
radio module and microcontroller (MCU) are powered off. The device periodically wakes up and activates internal
sensor devices (radar, magnetometer) for checking the change in parking status. In the current consumption snapshot
shown in Figure 14(right) sensor clearly wakes up every 60 seconds. If parking lot status changes (from free goes to
occupied), sensor device will send a LoRaWAN message over the radio to the gateway as can be seen in Figure 14(left),
which will result in increase of current consumption. The consumption of sensors and the MCU is slightly larger than
100 mA, along with LoRa transmission it goes over 150 mA. While MCU is in sleep state, the overall consumption
is around 0.01 A (Figure 15). In Libelium documentation, Smart Parking nodes have an expected lifetime up to ten
years of uninterrupted operation, which is assumed under normal circumstances and depending on settings of the smart
parking sensor device (night operation work regime, keep-alive packet transmission period, frequency of parking lot

11
 Figure 15: Consumption of Libelium Smart Parking sensor with Current Raider in sleep mode.

change status, etc.).

For example, Libelium smart parking sensors are equipped with built-in lithium-thionyl chloride (Li-SOCl2) bat-
teries with an overall capacity of 10.4 Ah. Assuming that device consumption during the sleep period (mA) is 0.01
mA, while device consumption during wake period is on average 100 mA, with 1 LoRaWAN keep-alive message sent
per hour and 5 seconds wake-up duration on average, an estimated battery lifetime will be around 2474.11 days, or
6.78 years. In this calculation it is assumed that the capacity is automatically derated by 15% from 10.4 Ah to account
for some self discharge. Note that periodic MCU wake-up (around every minute), as well as frequent parking lot
status changes and this LoRa radio communication will significantly reduce sensor lifetime duration, and they were
not taken into this preliminary calculation. Also, nighttime regime where LoRaWAN keep-alive messages can be sent
less frequently were also not taken into account in battery lifetime estimation.

5. Strategies for Optimization of Smart Parking Sensor Devices

To cope with the power consumption issues introduced in previous sections, in this section two strategies for the
optimization are proposed. First strategy shows how information on the received signal strength of the LPWA device
could serve as an indicator of the occupancy, where a drop in signal strength can be detected on the receiving side
when the vehicle is located above the sensor device. Second one proposes a novel device that compensates expensive,
mainly battery-operated and power hungry LPWA-based smart parking sensors.

5.1. RSSI as an Indicator of Occupancy

In this section analysis of signals from the LPWA devices installed in the parking lot, namely NB-IoT, Sigfox and
LoRaWAN, is presented. Arduino LPWA modules introduced in Section 3 were used that incorporate LoRa, Sigfox
and NB-IoT radio communication modules. Figure 16 shows the setup tested all of the above modules. To store the
data sent by the device, a server was created that implements Node-Red service, InfluxDB database and Grafana for
visualization of the collected data. In the following section architecture for every LPWA technology is described.

5.1.1. Technology used to analyze LPWA signals

As NB-IoT technology is based on existing LTE infrastructure from telecom operators, where it is necessary to
obtain a SIM card from a telecom operator supporting NB-IoT infrastructure to establish connection with it. In our
implementation T-Com operator was used for NB-IoT. Figure 17(up) shows the architecture of the NB-IoT implemen-
tation. As shown, MQTT protocol service was used to send data to our server. In particular, MKR NB 1500 device
periodically sent/published messages via MQTT protocol to a publicly available broker (mosquitto.com) to the topic
fesb/LPWA/rssi. On the server side, Node-RED subscribed to all messages on the previously mentioned topic
that reached the broker. Furthermore, all received messages from MQTT block were forwarded to the parser block,
12
 Figure 16: Test setup for the analysis of LPWA data composed of NB-IoT, Sigfox and LoRa devices.

Arduino MKR NB 1500 LTE base station MQTT broker

MQTT publish MQTT publish MQTT subscribe

Arduino MKRFOX 1200 Sigfox base station

HTTP HTTP

LoRaWAN Network and

Arduino MKR WAN 1300 base station Application broker

MQTT subscribe

Figure 17: Network architecture of (up) NB-IoT, (center) Sigfox and (down) LoRaWAN implementation.

that prepared message for further storage in InfluxDB, which can be further used for analysis and/or visualization in
Grafana.
Sigfox, similarly to NB-IoT, uses existing base stations provided by supporting operators. After registering the
device, Sigfox gives users access to their cloud system for further processing of messages received over base station.
The implementation architecture of Sigfox system is given in Figure 17(center). As can be seen, once the message
is sent over the radio from end device, base station forwards it to the Sigfox backend. Sigfox backend, on the other
hand, implements the callback functions that allow message forwarding to other services (e.g. email, IFTTT, thinger.io
etc.). HTTP integration was used to forward messages to ThingerIO cloud, that acts as bridge between Sigfox backend
and our server, since our server is not on a public IP address. Finally, a flow was created in Node-RED to capture
messages from ThingerIO. More specifically, first block periodically hits HTTP GET message to receive response
from ThingerIO cloud. The received message is forwarded towards report by exception (RBE) block that forwards
the message only if it differs from the previously received. This block is used to filter out duplicate messages received
at different time intervals. Next, the message is parsed and prepared for storage in InfluxDB.

13
 Finally, LoRaWAN in our example utilizes The Things Network cloud system that implements both Application
and Network server. Figure 17(down) shows the network architecture of our LoRaWAN system. In our example,
MKR WAN 1300 was used. As a LoRaWAN base station, an indoor base station close to the parking lot in our
university building was deployed. In our scenario, LoRaWAN implements the functionalities of The Things Network
base station, and forwards all received packets to the Network and Application server provided by The Things Network
(TTN). Flow was created in Node-RED on the server side where the first block establishes connection to The Things
Network, specifically, to the previously created application in TTN cloud that receives messages from the MKR WAN
end device.

5.1.2. Acquired data

This section presents preliminary results of the analysis of RSSI signals collected from the environment. The test
setup was designed in such a way that a parking lot was tested at a nearby parking lot and information of the RSSI was
collected in InfluxDB. Every device periodically sent messages over the radio. The signals were sent under conditions
with and without vehicle presence in the parking lot. In Figure 18 RSSI from Sigfox and LoRaWAN devices is
displayed. NB-IoT has proven to be extremely unreliable and was discarded from the observations. Under laboratory
conditions, the signal strength was 99 dBm, while at the parking lot the communication was unreliable and thus the
graph was omitted from further analysis. Interestingly, in situation where the vehicle is present in the parking lot,
RSSI is on average smaller than in the situation when the parking lot is vacant. This is more visible for LoRaWAN
technology than for Sigfox. These results are understandable given that the propagation effects and obstacles are less
expressed for the LoRaWAN since base station is located near the parking lot (30 m from the parking lot), compared
to the Sigfox base station placed at least 10 km from the parking lot. This indicates that the drop in signal strength
can serve as an indicator of vehicle presence, without any requirement for complex and expensive sensor devices.

5.2. Requirements for Developing a Novel Device

Performance restrictions of long range radios and other IoT sensing devices is caused by the limited energy source
(De Donno and Tarricone, 2013). Frequent transmissions require IoT node to be wake, sense the data and use power-
hungry radio interface of a long range. The batteries, including self-draining issue would last for a couple of years that
puts great restrictions for a usage scenario, since replacements would be a heavy task. In any case - energy source,
i.e. the batteries represent a major drawback for IoT-enabled environment. In order to compensate the needs, in this
section we propose a novel IoT device that should be:
• event driven and consume the energy peaks only during active state
• triggered to wake-up once the environmental change happen. This action should be near zero-energy procedure
• multiple-source energy harvesting enabled

-100 -105
Occupied lot Free lot Occupied lot Occupied lot Free lot Occupied lot

-110
-110

-120
RSSI

RSSI

-115
-130

-120
-140

-150 -125
9:40 9:45 9:50 9:55 10:00 10:05 10:10 9:40 9:45 9:50 9:55 10:00 10:05 10:10
Time Time

Figure 18: Analyzed RSSI from LPWA devices (left) Sigfox, (right) LoRaWAN.

14
 SOLAR CELL

CONTROL
UNIT

to IoT Device
1.8 V
to IoT node

UHF RFID Interface

Figure 19: Wake-up trigger and Energy harvesting architecture

• energy efficient in the way to match device duty cycle

• long distance. Coverage of many low-range readers cause heavy infrastructure needs, while backscattering
appears to be faulty and the technology itself is not yet mature enough
• compatible for providing the information for training in machine learning techniques.
The aforementioned drawbacks could be partially compensated by accumulating the available energy into the
storage and enable the system for later use. However, the most important part of the architecture, is the ”Control
unit” depicted in Figure 19. The aim of the device is to use energy source inputs, when above the threshold to a
wake-up the power hungry circuitry of a generic IoT node. The control unit will deliver the energy to energy storage
element, developed by means of rechargeable batteries or capacitors. The voltage of IoT device is then regulated by
Low Dropout (LDO) voltage regulator. It is important to note that this system should be able to match the power
requirements of related duty cycle and the node itself. That case would result with a sustainable device that delivers
data once the energy is present, and is even being more efficient as the wake-ups are triggered by the environmental
change.

6. Related Work

A plethora of solutions exists that utilize sensors for the detection of parking place. Infrared sensors implemented
in the context of smart parking solutions can be active and passive. For example, passive infrared sensors work as
detectors of heat from surrounding objects, such as humans passing in the vicinity of the sensor device, i.e. parking
lot (Someswar et al., 2017; Song et al., 2008). These types of sensor device are normally combined with other sensors
in terms of vehicle detection. For example, in (Larisis et al., 2012) combination of passive infrared sensor (PIR) and
magnetometer is employed for vehicle presence detection. Using the active infrared sensor device an estimation of
distance is measured between the sensor device and a surrounding object in front of them (Moguel et al., 2014).
On the other hand, an ultrasound sensor measures ultrasound round-trip reflected from an object positioned in
front of them. This way, an ultrasound sensor device is more suitable for outdoor environments as it estimates the
distance between obstacles (Lee et al., 2008; Coric and Gruteser, 2013). In addition, these sensor devices can be
useful to calculate vehicle speed as well as the number of vehicles (Kianpisheh et al., 2012).
15
 Accelerometers are also applied in smart parking solutions for vehicle detection as they measure ground vibrations
for detecting the vehicle approaching the parking spot. Merged with other sensor devices, such as optical sensors, a
better estimation of vehicle presence could be obtained (Gharaibeh et al., 2017). In (Jeffrey et al., 2012) ambient light
sensor is used to detect the presence of the vehicle.
The majority of todays smart parking sensor devices for the sensing of vehicle at the parking spot use magnetome-
ters that detect metal within the change of the magnetic field when an object is placed within their vicinity 7 . In (Arab
and Nadeem, 2017) smartphone magnetometers are utilized for detecting vehicles placed at the parking lot.
Cameras can also be used for parking occupancy detection as they require more complex image processing in real-
time (Nguyen and Nguyen, 2014; Sevillano et al., 2014). Using spatiotemporal video analysis cameras are also used to
detect the presence of vehicles parked in street lanes (Wahyono and Jo, 2017). (Bulan et al., 2013) introduced camera-
based system for the detection of parkging occupancy that operates in real-time. In (Bong et al., 2011) introduce Car
Park Occupancy Information System (COINS) that uses image processing techniques for occupied/available parking
lot. Similarly, (Almeida et al., 2015) use a large dataset of figures with Local Binary Patterns and Local Phase
Quantization, along with Support Vector Machine classifier for parking lot vacancy detection. Other sensor devices,
such as piezoelectric sensors and/or inductive loops are used in roads for detecting a passing vehicle (Kotb et al.,
2017). At the end, RFID as a technology can be also used in smart parking systems, mostly for payments (Abdullah
et al., 2013; Zhao, 2012). In (Bagula et al., 2015), RFID is used as a replacement for paper-based ticketing.
After detecting the occupancy or reading the sensor, smart parking sensor nodes transmit information to the cen-
tralized systems. Usually, a wireless technology is utilized in smart parking systems, ranging from short to wide range
radio technologies such as: Wi-Fi ZigBee/DigiMesh, Bluetooth/BLE, or long-range, like LoRa, Sigfox, NB-IoT/LTE-
M and Ingenu (de Carvalho Silva et al., 2017; Raza et al., 2017). Compared to short range radio technology such as
WiFi and/or ZigBee, where multihop radio communication channel is employed for information transmission to the
base station, long range radio systems found in LoRa or Sigfox use single hop communication channel which in many
cases makes them suitable for smart parking solutions.

7. Conclusion

In the near future almost 70 percent of humans will live in urban areas. Taking into account that in large cities,
even today an average driver has to spend a large portion of time to find a free parking lot, in the overall it increases
street congestion which rises fuel consumption that contributes to air pollution. Smart and Sustainable Mobility with
the enormous advancement of IoT presents one of the visions of Smart Cities. In this vein, the development of IoT-
based Smart Parking solutions, especially in overpopulated areas will improve the quality of life and result in cleaner
and greener environment.
This paper gives an extension to state-of-the art by a systematic, in-depth overview of technologies used for
smart parking detection realization focusing on performances of most popular sensors used nowadays as well as
power requirements analysis for supporting the sensing infrastructure. Further, an analysis on the performance of the
photodiode, Light-dependent resistor (LDR), Infrared LED, Ultrasonic, LiDAR and magnetic sensors were provided.
From the power consumption perspective photodiode should be the most appropriate vehicle detection sensor, however
it lacks detection accuracy. Therefore, the most appropriate results in terms of both consumption and accuracy would
be 3-axis magnetic sensor, which is followed by ultrasonic sensor. In addition, performances of Low Power Wide Area
radio communication technologies, such as LoRa, NB-IoT and Sigfox were analyzed, both in terms of availability and
consumption, which are commonly found in today’s smart parking detection systems. LoRaWAN was shown to be
the best LPWA technology since it is infrastructure independent, and power consumption is by far smaller compared
to other LPWA technologies (Sigfox and NB-IoT).
To cope with issues resulting from potential high power consumption of battery operated devices, two strategies
for optimization were proposed. First solution uses a drop in signal strength on the receiving side which can serve
as an indicator of parking space occupancy once the vehicle is located above the sensor device. More particularly,
RSSI signals collected from the environment using LoRaWAN, Sigfox and NB-IoT technology indicate that a drop in

7 ”China Parking,” www.chinaparking.org, ”SmartSantander,” www.smartsantander.eu, ”LA ExpressPark,” www.

laexpresspark.org

16
signal strength can indeed serve as vehicle present at the parking lot. A second solution proposed a novel device that
compensates expensive, mainly battery-operated and power hungry LPWA-based smart parking sensors. In particular,
a big picture on novel architecture for harvesting the surrounding energy and using the same for circuitry wake-up
therefore saving the energy is proposed.

Acknowledgment

This work was partially supported by the Croatian Science Foundation under the project “Internet of Things:
Research and Applications”, UIP-2017-05-4206; by FCT/MCTES through national funds and when applicable co-
funded EU funds under the Project UIDB/EEA/50008/2020; and by Brazilian National Council for Research and
Development (CNPq) via Grant No. 309335/2017-5.

References
Abdullah, S., Ismail, W., Halim, Z., Zulkifli, C., 2013. Integrating Zigbee-based mesh network with embedded passive and active rfid for production
management automation, in: IEEE Intl Conf on RFID-Technologies and Applications.
Al-Turjman, F., 2018. Mobile Couriers’ selection for the Smart-grid in Smart-cities’ Pervasive Sensing. Future Generation Comp. Syst. 82,
327–341.
Al-Turjman, F., Malekloo, A., 2019. Smart Parking in IoT-enabled Cities: A Survey. Sustainable Cities and Society 49, 101608.
Almeida, P., Soares de Oliveira, L., Jr, A., Jr, E., Koerich, A., 2015. PKLot - A Robust Dataset for Parking Lot Classification. Expert Systems with
Applications 42. doi:10.1016/j.eswa.2015.02.009.
Arab, M., Nadeem, T., 2017. MagnoPark - Locating On-Street Parking Spaces Using Magnetometer-Based Pedestrians’ Smartphones, in: 2017
14th Annual IEEE International Conference on Sensing, Communication, and Networking (SECON), pp. 1–9.
Bagula, A., Castelli, L., Zennaro, M., 2015. On the design of smart parking networks in the smart cities: An optimal sensor placement model 15.
Bibri, S.E., 2018. The IoT for smart sustainable cities of the future: An analytical framework for sensor-based big data applications for environ-
mental sustainability. Sustainable Cities and Society 38, 230–253.
Bong, D., Ting, K., Lai, K., 2011. Integrated approach in the design of car park occupancy information system (COINS). IAENG Intl J. of
Computer Science 35.
Bulan, O., Loce, R.P., Wu, W., Wang, Y., Bernal, E.A., Fan, Z., 2013. Video-based real-time on-street parking occupancy detection system. J. of
Electronic Imaging 22.
Centenaro, M., Vangelista, L., Zanella, A., Zorzi, M., 2016. Long-Range Communications in Unlicensed Bands: the Rising Stars in the IoT and
Smart City Scenarios. IEEE Wireless Communications 23.
Coric, V., Gruteser, M., 2013. Crowdsensing maps of on-street parking spaces, in: IEEE Intl Conf on Distributed Computing in Sensor Systems.
Cruz, M.A.A., Rodrigues, J.J.P.C., Gomes, G., Almeida, P., Rabelo, R.A.L., Kumar, N., Mumtaz, S., 2019. An IoT-based Solution for Smart
Parking, in: International Conference on Computing Communication and Cyber Security (IC4S 2019).
de Carvalho Silva, J., Rodrigues, J.J.P.C., Alberti, A.M., Solic, P., Aquino, A.L.L., 2017. Lorawan — a low power wan protocol for internet of
things: A review and opportunities, in: 2017 2nd International Multidisciplinary Conference on Computer and Energy Science (SpliTech), pp.
1–6.
De Donno, D., C.L., Tarricone, L., 2013. A battery-assisted sensor-enhanced RFID tag enabling heterogeneous wireless sensor networks. IEEE
Sensors Journal 14, 1048–1055.
Geng, Y., Cassandras, C.G., 2013. New “smart parking” system based on resource allocation and reservations. IEEE Transactions on Intelligent
Transportation Systems 14, 1129–1139.
Gharaibeh, A., Salahuddin, M.A., Hussini, S.J., Khreishah, A., Khalil, I., Guizani, M., Al-Fuqaha, A., 2017. Smart cities: A survey on data
management, security, and enabling technologies. IEEE Communications Surveys Tutorials 19, 2456–2501.
Jeffrey, J., Patil, R., Narahari, S., Didagi, Y., Bapat, J., Das, D., 2012. Smart parking system using wireless sensor networks, in: Intl Conf on Sensor
Technologies and Applications.
Kianpisheh, A., Mustaffa, N., Limtrairut, P., Keikhosrokiani, P., 2012. Smart parking system (sps) architecture using ultrasonic detector. Interna-
tional Journal of Software Engineering Applications 6(3), 55–58.
Kotb, A.O., Shen, Y., Huang, Y., 2017. Smart parking guidance, monitoring and reservations: A review. IEEE Intelligent Transportation Systems
Magazine 9, 6–16.
Kotb, A.O., Shen, Y., Zhu, X., Huang, Y., 2016. iparker—a new smart car-parking system based on dynamic resource allocation and pricing. IEEE
Transactions on Intelligent Transportation Systems 17, 2637–2647.
Larisis, N., Perlepes, L., Kikiras, P., Stamoulis, G., 2012. U-park: Parking management system based on wireless sensor network technology, in:
Intl Conf on Sensor Technologies and Applications.
Lee, S., Yoon, D., Ghosh, A., 2008. Intelligent parking lot application using wireless sensor networks, in: Intl Symp on Collaborative Technologies
and Systems.
Lin, T., Rivano, H., Le Mouël, F., 2017. A survey of smart parking solutions. IEEE Transactions on Intelligent Transportation Systems 18,
3229–3253. doi:10.1109/TITS.2017.2685143.
Moguel, E., Preciado, M., Preciado, J., 2014. Smart parking campus: An example of integrating different parking sensing solutions into a single
scalable system, in: ERCIM News Smart Cities.

17
 Nguyen, T., Nguyen, C., 2014. An approach for building an intelligent parking support system, in: Symp on Information and Communication
Technology.
Nizetic, S., Djilali, N., Papadopoulos, A., J., J.R., 2019. Smart technologies for promotion of energy efficiency, utilization of sustainable resources
and waste management. Journal of Cleaner Production 231, 565–591.
Perković, T., Damjanović, S., Šolić, P., Patrono, L., Rodrigues, J.J., 2019. Meeting Challenges in IoT: Sensing, Energy Efficiency and the
Implementation, in: Fourth International Congress on Information and Communication Technology in concurrent with ICT Excellence Awards
(ICICT 2019).
Pham, T.N., Tsai, M., Nguyen, D.B., Dow, C., Deng, D., 2015. A cloud-based smart-parking system based on internet-of-things technologies.
IEEE Access 3, 1581–1591.
Raza, U., Kulkarni, P., Sooriyabandara, M., 2017. Low power wide area networks: An overview. IEEE Communications Surveys Tutorials 19,
855–873.
Sevillano, X., Marmol, E., Fernandez-Arguedas, V., 2014. Towards smart traffic management systems: Vacant on-street parking spot detection
based on video analytics, in: Intl Conf on Information Fusion.
Solic, P., Colella, R., Catarinucci, L., Perkovic, T., Patrono, L., 2019. Proof of Presence: Novel Vehicle Detection System. IEEE Wireless
Communications URL: http://dx.doi.org/10.1109/MWC.001.1900133.
Someswar, G.M., Dayananda, R.B., Anupama, S., Priyadarshini, J., Shariff, A.A., 2017. Design & development of an autonomic integrated car
parking system. Compusoft 6(3), 2309.
Song, B., Choi, H., Lee, H.S., 2008. Surveillance Tracking System Using Passive Infrared Motion Sensors in Wireless Sensor Network, in: 2008
International Conference on Information Networking, pp. 1–5.
Wahyono, Jo, K., 2017. Cumulative dual foreground differences for illegally parked vehicles detection. IEEE Transactions on Industrial Informatics
13, 2464–2473. doi:10.1109/TII.2017.2665584.
Zanella, A., Bui, N., Castellani, A., Vangelista, L., Zorzi, M., 2014. Internet of things for smart cities. IEEE Internet of Things Journal 1, 22–32.
Zhao, W., 2012. Parking management system with rfid and sensor networks 198-199.
Šolić, P., Perković, T., Konsa, T., Zargariasl, H., Patrono, L., 2019. Smart parking sensor performance evaluation, in: 2019 International Conference
on Software, Telecommunications and Computer Networks (SoftCOM), pp. 1–6.

18

View publication stats