Padova Smart City: an Urban Internet of Things

Experimentation
Angelo Cenedese, Member, IEEE, Andrea Zanella, Senior Member, IEEE, Lorenzo Vangelista, Senior
Member, IEEE, Michele Zorzi, Fellow, IEEE

AbstractSmart City is a powerful paradigm that applies

the most advanced communication technologies to urban environments, with the final aim of enhancing the quality of life
in cities and provide a wide set of value-added services to
both citizens and administration. A fundamental step towards
the practical realization of the Smart City concept consists in
the development of a communication infrastructure capable of
collecting data from a large variety of different devices in a
mostly uniform and seamless manner, according to the Internet
of Things (IoT) paradigm. While the scientific and commercial
interest in IoT has been constantly growing in the last years,
practical experimentation of IoT systems has just begun. In this
paper, we present and discuss the Padova Smart City system,
an experimental realization of an urban IoT system designed
within the Smart City framework and deployed in the city of
Padova, Italy. We describe the system architecture and discuss
the fundamental technical choices at the base of the project.
Then, we analyze the data collected by the system and show how
simple data processing techniques can be used to gain insights on
the functioning of the monitored system, public traffic lighting
in our specific case, as well as other information concerning the
urban environment.
Index TermsSmart Cities, Test-bed and Trials, Sensor System Integration, Network Architecture, Service Functions and
Management, EXI, CoAP, 6LoWPAN

I. I NTRODUCTION
Generally speaking, the concept of Smart City consists
of exploiting the modern Information and Communication
Technologies (ICT) in operating the public affairs. The aim
is to make a better use of the public resources, increase the
quality of the services offered to the citizens and, in turn,
the quality of life in the urban areas, while reducing the
operational costs of the public administrations. Behind this
vision, analysts foresee a potentially huge market, which is
estimated at hundreds of billion dollars by 2020 [1]. To unleash
the potential of such a market, however, the Smart City concept
shall be adopted by various actors, as key industries, service
providers, and public administrations, and applied in a synergic
manner to different sectors, thus taking the flavor of Smart
Governance, Smart Mobility, Smart Utilities, Smart Buildings,
and Smart Environment.
Angelo Cenedese, Andrea Zanella, Lorenzo Vangelista, and Michele Zorzi
are with the Department of Information Engineering, University of Padova,
Via Gradenigo 6/B - 35131, Padova, Italy. They are also affiliated with
Human Inspired Technologies (HIT) Research Center, University of Padova,
and Andrea Zanella, Lorenzo Vangelista, and Michele Zorzi are also affiliated
with Consorzio Ferrara Ricerche (CFR), Via Saragat, 1 - 44122, Ferrara, Italy.
Michele Zorzi and Lorenzo Vangelista are founding members of Patavina
Technologies s.r.l., Via Venezia 59/8 - 35131, Padova, Italy.
Contact authors e-mail: [email protected].

In this complex scenario, public administrations may take

the pivotal role of early adopter for those services that combines social utility with very clear return on investment, such
as smart lighting, smart parking, smart buildings, which can
then open the way for other value-added services [2]. To this
end, however, these initial services need to be developed within
a more general framework that allows for seamless integration
of other services later.
In this respect, the Internet of Things (IoT) paradigm can
become the building block to realize a unified urban-scale
ICT platform, thus unleashing the potential of the Smart City
vision [3], [4]. The IoT is indeed a recent communication
paradigm that aims at annexing into the Internet any kind
of object, provided it is equipped with a micro-controller, a
communication transceiver, and a suitable protocol stack [5].
Hence, the IoT can enable easy access and interaction with a
wide variety of devices such as, for instance, home appliances,
surveillance cameras, monitoring sensors, actuators, displays,
vehicles, and so on. The potentially enormous amount and
variety of data generated by such objects will then foster
the development of new services to citizens, companies, and
public administrations [6]. Therefore, the adoption of the IoT
paradigm in a Smart City scenario is very attractive to public
administrations, which may become the promoters for the
adoption of the IoT paradigm on a wider scale [7].
The objective of this paper is to describe the Padova Smart
City (PSC) project, a practical implementation of an urban
IoT realized in the city of Padova, in Italy. The project is the
result of the collaboration between public and private parties,
such as: the municipality of Padova, which has sponsored the
project; the Department of Information Engineering and the
Human Inspired Technologies research centre of the University
of Padova, which have provided the feasibility analysis of the
project and the data post-processing; Patavina Technologies
s.r.l.1 , a spin-off of the University of Padova specialized in the
development of innovative IoT solutions, which has developed
the control software of the employed IoT nodes manufactured
by Smart Future s.r.l..2
The target application consists of a system for monitoring
the public street lighting by means of wireless nodes, equipped
with different kinds of sensors, placed on street light poles
and connected to the Internet through a gateway unit. Besides
light intensity measurements at each post, which can be used
to check the correct operation of the public lighting system,
1 http://patavinatech.com/
2 http://www.smartfuture.eu

Fig. 1.

System architecture of Padova Smart City.

this system makes it possible to collect other interesting

environmental parameters, such as CO level, air temperature
and humidity, vibrations, noise, and so on. Even if this system
is a simple application of the IoT concept, it still involves a
number of different devices and link layer technologies, thus
being representative of most of the critical issues that need to
be taken care of when designing an urban IoT. Furthermore,
we present some experimental data collected by the PSC
system and show with a couple of examples how the simple
processing of raw data can provide interesting information
related to the public lighting system and the level of pollution
in the air.
The rest of this paper is organized as follows. Sec. II describes the PSC components and the web service architecture
used to collect and export the data. Sec. III reports some
selected measurements to exemplify the type of data that can
be collected with the system and the kind of analysis and
inference that can be performed on them. Finally, in Sec. IV
we draw the conclusions.
II. PSC SYSTEM ARCHITECTURE
The conceptual architecture of the Padova Smart City system, sketched in Fig. 1, has been designed according to the
web-based framework discussed in [7] and partly reported here
for the readers convenience.
The system consists of a few IoT sensor nodes placed on
streetlight poles and connected to the network of the city
municipality by means of a gateway. Each IoT node is geographically localized, so that IoT data can be enhanced with
context information. The nodes are equipped with photometer
sensors that directly measure the intensity of the light emitted
by the lamps (or, actually, by any source whose light reaches
the sensor) at regular time intervals or upon request. The
wireless IoT nodes are also equipped with temperature and
humidity sensors, which provide data concerning weather conditions, and one node is also equipped with a benzene (C6 H6 )
sensor, which monitors air quality. IoT nodes are powered by
small batteries, so that each unit is self-contained and can be

easily placed in any location. The only exception is the node

equipped with the benzene sensor that, requiring much higher
power supply, has been located in the only position where a
DC connector was available. The sensor nodes are packaged
in a transparent plastic shield that protects the electronic parts
from atmospheric phenomena, while permitting the circulation
of air and light for the correct measurement of humidity,
temperature, and light intensity.
The PSC project adopts IETF standards which are open
and royalty-free, thus satisfying the Italian national directive
that requires the use of open source software in government
and in public offices. The IETF standards for IoT foresee
a web service architecture for IoT services, which has been
largely investigated in the literature [7]. This approach makes it
possible to develop flexible IoT services that can easily interact
with other web services through the adoption the Representational State Transfer (ReST) paradigm [8]. This paradigm,
indeed, guarantees strong similarities in the structure of IoT
and traditional web services, thus promoting the adoption of
IoT by both end users and service developers.
The web service approach requires the deployment of suitable protocol layers in the different elements of the network, as
shown in the protocol stacks depicted in Fig. 1 besides the key
elements of the system architecture. In particular, we note that
the common de-facto standards for Internet communications,
such as HTTP, IPv4/v6, and Ethernet, are replaced in the
resource-constrained devices (as the sensor nodes) by their IoT
counterparts, i.e., Constrained Application Protocol (CoAP),
IPv6, and 6LoWPAN, whose role is better described in the
following.
A. HTTP-CoAP Proxy
In today Internet, most of the data traffic is carried by HTTP
over TCP. Therefore, HTTP seems to be the most natural
choice for data transfer in IoT as well. Unfortunately, the
straight application of HTTP in this context is prevented by the
computational and processing constraints of some IoT nodes.
Another problem is represented by the TCP transport protocol

that yields poor performance in lossy environments. The

CoAP protocol [9] overcomes these difficulties by proposing
a binary format transported over UDP, handling only the
retransmissions strictly required to provide a reliable service.
Moreover, CoAP can easily interoperate with HTTP because
it supports the ReST methods of HTTP (GET, PUT, POST,
and DELETE), and the response codes of the two protocols
are in one-to-one correspondence.
In order to enable easy interaction with the PSC system
from any common Internet host, we deployed an HTTP-CoAP
intermediary, also known as cross proxy, at the wireless sensor
network (WSN) gateway. The HTTP-CoAP proxy enables
transparent communication with CoAP devices. The proxy
logic can be extended to better support monitoring applications
and limit the amount of traffic injected into the IoT peripheral
network. For instance, it is possible to specify a list of
resources that need to be monitored, so that the server can
autonomously update the entries in a cache related to those
devices. This mechanism can be supported by two different
approaches: (i) by polling the selected resource proactively,
thus enabling the implementation of traffic shaping techniques
at the proxy or at the gateway, and (ii) by subscribing to the
selected resource using the observe functionality of CoAP,
thus enabling the server on the node to send the updates only
when the value measured by the sensor falls outside a certain
range. The PSC project adopts the first option, mainly because of its implementation simplicity. Furthermore, while the
gateway provides the HTTP-CoAP mapping, at the moment
the peripheral nodes do not support CoAP service yet, so that
data collection is performed by the WSN gateway through a
dedicated application. However, the software architecture has
been designed to allow for a smooth migration of the system
to a full web-based framework as soon as the CoAP service
will be implemented in the sensor nodes.
B. Network layer
For what concerns the network layer, the PSC system employs a mix of IPv4 and IPv6 protocols [10]. More specifically,
the IoT nodes are assigned 128-bit long IPv6 addresses that
guarantee global uniqueness. On the other hand, the long
address field introduces overheads that are not compatible with
the scarce capabilities of constrained nodes. For this reason,
the IoT nodes make use of the 6LoWPAN protocol [11], [12],
which is an established compression format for IPv6 and UDP
headers over low-power constrained networks. The gateway
transparently translates any IPv6 packet intended for a node in
the 6LoWPAN network into a packet with 6LoWPAN header
compression format, and operates the inverse translation in the
opposite direction.
The interaction with IPv4-only hosts is obtained by means
of an IPv4/IPv6 Port Address Translation (v4/v6 PAT). This
method maps arbitrary pairs of IPv4 addresses and TCP/UDP
ports into IPv6 addresses and TCP/UDP ports. It resembles
the classical Network Address and Port Translation (NAPT)
service currently supported in many LANs to provide Internet
access to a number of hosts in a private network by sharing
a common public IPv4 address. The same technique can be

used to map multiple IPv6 addresses into a single IPv4 public

address, which allows the forwarding of the datagrams in the
IPv4 network and its correct management at IPv4-only hosts.
We note that this approach raises a scalability problem due
to the limited number of available TCP/UDP ports (65535).
Nonetheless, it works perfectly to access a relatively small
island of IoT nodes that can be aggregated by a single gateway,
as in the case of the PSC system.
C. Link layer technology
The IoT nodes are equipped with a CC2420 transceiver,
manufactured by Texas Instruments, that implements the IEEE
802.15.4 standard [13], [14]. Routing functionalities are provided by the IPv6 Routing Protocol for Low power and Lossy
Networks (RPL) [15]. Nodes collectively deliver their data
to the gateway, which represents the single point of contact
for the external nodes. The gateway hence plays the role of
6LoWPAN border router and RPL root node. Furthermore,
since sensor nodes do not support CoAP services, the gateway
also operates as the sink node for the sensor cloud, collecting
all the data that need to be exported to the backend services.
The connection to the backend services is provided by optical
fiber through a virtual private network (VPN) connection.
D. Database server
The database server is realized within the WSN Gateway,
which hence represents a plug-and-play module that provides a
transparent interface with the peripheral nodes. At the moment,
the state of the resources that need to be stored in the database
are collected by the WSN Gateway by means of a dedicated
application. In the future, however, the WSN gateway will
be replaced by the HTTP-CoAP proxy server, which will take
care of retrieving the required data from the proper source. The
data stored in the database can be accessed through traditional
web programming technologies and visualized in the form of
a web site, or exported in any open data format using dynamic
web programming languages.
Such a system can be successively extended to include other
types of IoT nodes or clouds of IoT nodes, provided that
each IoT peripheral system supports an HTTP-based interface,
which makes it possible to interact with it in an open, standard,
and technology independent manner.
III. A NALYSIS OF PSC COLLECTED DATA
Before being interpreted to retrieve high-level information,
the data collected by the PSC system are processed by
applying a moving average filter over a time window of one
hour (approximately 10 readings of temperature, humidity and
light, and 120 readings of the benzene sensor, whose sampling
rate is larger since the node is powered by the grid). The data
are then analyzed to detect anomalies and events that may
occur in the monitored area, to profile the users behaviors,
and to gain insight into the use of city resources, which may
serve to optimize or create services for the community.
As an example, we report the data measured by the set of
eight sensors positioned along the high-traffic road as shown
in Fig. 2.

S08

S06

S07

S02

3500

3000

Value [lux]

S05

4000

S04
S03
S02
S01
WSN gateway
benzene sensor

2500

2000

1500

1000

500

10/29

11/05

11/12

11/19
11/26
Time [days]

12/03

12/10

(a)

Fig. 2. Sample deployment of nodes for PSC: nodes position is indicated

by the green markers.
4000

S07

3500

3000

Value [lux]

In Fig. 3 we show the time series of the light measurement

signal collected by sensors S02 (Fig. 3(a)) and S07 (Fig. 3(b)).
In general, it is possible to observe the regular pattern of the
light measurements, corresponding to day and night periods.
In particular, at daytime the measure reaches the saturation
value, while during nighttime the values are more irregular,
also due to the reflections produced by vehicle lights. The
first example, in the context of anomaly detection, is summarized in Fig. 4, where the light signals collected by all
sensors are processed to obtain some aggregate values. More
specifically, a variance analysis of the whole time-series is
shown in Fig. 4(a), which clearly highlights the presence of
two clusters of measurements, respectively obtained by sensors
{S01, S02, S03, S04, S06} and {S05, S07, S08}, indicated
by the red dashed circles. This result is confirmed by the time
evolution of the daily variances shown in Fig. 4(b) that also
shows the malfunctioning of S02 (dark yellow line) during the
first 25 days of the experimental campaign (see also Fig. 3(a)).
A deeper understanding of the phenomena can be obtained
by observing the time evolution of the variances restricted
to the night hours, as reported in Fig. 4(c). The night
period variance supports a further better classification: two
sensors are grouped together, namely {S05, S08} (purple and
cyan lines); the four referred to {S01, S03, S04, S06} signals
have low and similar variance (red, light and dark green,
light blue); finally sensor S07 (blue line) shows a more
erratic behavior. The higher measurement stability provided
by {S01, S03, S04, S06} with respect to that of {S05, S08}
suggests a flickering illumination given by the latter set.
The clustering of the light sensors suggested by this last
analysis is also confirmed by the visual comparison of the
time series of the signals measured by the different sensors,
as reported in Fig. 5 for the sensors in the first two clusters,
i.e., {S01, S03, S04, S06} and {S05, S08}.
The second example is reported in Fig. 6 (at the end of
the paper) and considers the benzene measurements. In the
figure, the grey bars indicate Sundays and the benzene signal
is shown together with the humidity evolution. Interestingly,

12/17

2500

2000

1500

1000

500
10/29

11/05

11/12

11/19
11/26
Time [days]

12/03

12/10

12/17

(b)
Fig. 3. Example of data collected by Padova Smart City: measured light
signal from sensor S02 (a) and sensor S07 (b).

over a longer horizon (10 weeks), the measurement time-series

shows a growing trend starting at the end of November that
relates well with the weather conditions shown in the subplot
Fig. 6(a) specifically with the rainfall and the average wind
strength (retrieved from the local weather station database). In
the absence of rainfall and in the presence of air stagnation,
the pollution levels indicated by the benzene measurements
tend to increase (the last three weeks reported in the plots are
particularly meaningful in this sense). In addition, these values
are also linked to an increment in the citizens mobility due
to Christmas shopping.
On a shorter timescale, it is possible to study the behavior
during daytime/nighttime, which is useful to optimize the
traffic flows. The two cases are presented: Fig. 6(b) remarks
the difference between a vacation day (Sunday) and a working
day (Monday), showing for the former a generally lower
pollution level without the peak that characterizes Mondays
early morning. A different situation is proposed in Fig. 6(c),
where the general behavior for Sunday is similar to the

x 10

Signal variance

Signal variance

3.2
3
2.8
2.6

10

S01
S02
S03
S04
S05
S06
S07
S08

10
Signal variance

3.4

10

S01
S02
S03
S04
S05
S06
S07
S08

10
2.4
2.2

S01 S02 S03 S04 S05 S06 S07 S08

Sensor

(a)

10
10/29 11/05 11/12 11/19 11/26 12/03 12/10 12/17
Time [days]

10
10/29 11/05 11/12 11/19 11/26 12/03 12/10 12/17
Time [days]

(b)

(c)

Fig. 4. Example of data collected by Padova Smart City: measured light signal. (a)-(b)-(c): signal variances in 70 days, 1 days, and in night period,
respectively.

previous example, but Monday signals show a completely

different evolution: a peak in the humidity is followed by a
rapid decrease, and correspondently the benzene signal drops
and then remains limited. The scenario is consistent with an
occurring storm that initially causes congestion in the road
traffic and, in turn, a peak of benzene in the air, and then
leaves a lower pollution level.

3000

Value [lux]

In this paper we presented Padova Smart City, a pilot

implementation of urban IoT within a Smart City framework.
We illustrated the system architecture, which adheres to the
web service paradigm, and described the adopted protocols
that satisfy the requirements for standardized and open source
solutions raised by the municipality that commissioned the
project. The system has been designed to allow for future
extension to include other types of IoT nodes or clouds of IoT
nodes, provided that each IoT peripheral system supports an
HTTP-based interface, which makes it possible to interact with
it in an open, standard, and technology-independent manner.
As an example of the possible utilizations of the data collected
by such a system, we also reported some snapshots of sensor
signals, namely light, humidity and benzene level in the air,
and show how a several considerations and inferences can be
drawn from the data through extremely simple data processing
techniques. As a future work, we plan to couple the sensor data
with location information provided by the GIS database and
with other data that are collected by the municipality using
dedicated systems (e.g., traffic intensity, parking occupancy,
weather conditions, and so on) and to apply more sophisticated data analysis techniques to unravel correlations among
different signals and get refined information from the raw data.

S01
S03
S04
S06

3500

2500

2000

1500

1000

500
10/29

11/05

11/12

11/19
11/26
Time [days]

12/03

12/10

12/17

(a)

4000

S05
S08

3500

3000

Value [lux]

IV. C ONCLUSIONS

4000

2500

2000

1500

1000

500

ACKNOWLEDGEMENT
The authors would like to thank the municipality of Padova
(Italy), and Eng. Alberto Cor in particular, for the support in
the realization of the Padova Smart City project. The authors
are also grateful to the engineers of Patavina Technologies
s.r.l. (http://patavinatech.com/) for their invaluable support in
deploying the system and in providing experimental data and
technical documentation concerning the Padova Smart City

10/29

11/05

11/12

11/19
11/26
Time [days]

12/03

12/10

12/17

(b)
Fig. 5. Example of data collected by Padova Smart City: measured light
signal. Signal time-series with sensors clustering according to the variance
analysis of Fig. 4(c).

25

60

20

50

15

40

10

30

20
10/13

10/20

10/27

11/03

11/10

11/17

11/24

12/01

12/08

0
12/22

12/15

Fallen Rain [mm]

20

10
30
15
0
10/13

10/20

10/27

11/03

11/10

11/17
Time [days]

11/24

12/01

12/08

Benzene [ g/m3]

30

70

0
12/22

12/15

Average Wind Strength [km/h]

Relative Humidity [%]

80

80

30

70

25

70

25

60

20

60

20

50

15

50

15

40

10

40

10

30

30

0
h.24

20

20

h.06

h.12

h.18

h.24
h.06
Time [hours]

h.12

h.18

(b)

h.06

h.12

h.18

h.24
h.06
Time [hours]

h.12

h.18

Benzene [ g/m3]

30

Relative Humidity [%]

80

Benzene [ g/m3]

Relative Humidity [%]

(a)

0
h.24

(c)

Fig. 6. Example of data collected by Padova Smart City: measured benzene signal (lower green line), plotted against the relative humidity signal (upper blue
line), and compared with the weather conditions related to rainfall (discontinuous blue line) and wind (continuous green line). (a): whole time-series analysis;
(b)-(c): daily analysis. The grey shadowing indicates Sundays.

project. This work has been supported in part by the European

Commission through the FP7 EU project Symbiotic Wireless
Autonomous Powered system (SWAP, G.A. no. 251557, http:
//www.fp7-swap.eu/).
R EFERENCES
[1] Pike Research on Smart Cities. [Online]. Available: http://www.
pikeresearch.com/research/smart-cities.
[2] N. Walravens, P. Ballon, Platform business models for smart cities:
from control and value to governance and public value IEEE Communications Magazine, vol. 51, no. 6, pp. 72-79, Jun. 2013.
[3] J. M. Hernndez-Muoz, J. B. Vercher, L. Muoz, J. A. Galache, M.
Presser, L. A. Hernndez Gmez, and J. Pettersson, Smart Cities at the
Forefront of the Future Internet, The Future Internet, Lecture Notes in
Computer Science, vol. 6656, pp. 447-462, 2011.
[4] C. E. A. Mulligan, M. Olsson, Architectural implications of smart city
business models: an evolutionary perspective IEEE Communications
Magazine, vol. 51, no. 6, pp. 80-85, Jun. 2013.
[5] L. Atzori, A. Iera, and G. Morabito, The internet of things: A survey,
Computer Networks, vol. 54, no. 15, pp. 2787-2805, 2010.
[6] D. Cuff, M. Hansen, and J. Kang, Urban sensing: out of the woods.
Communications of the ACM vol. 51, no. 3, pp. 24-33, Mar. 2008.

[7] A. Zanella, N. Bui, A. Castellani, L. Vangelista, M. Zorzi, Internet of

Things for Smart Cities, IEEE Internet of Things Journal, vol.1, no.1,
pp.22,32, Feb. 2014.
[8] R. Fielding, Representational state transfer, Architectural Styles and
the Design of Network-based Software Architecture, pp. 76-85, 2000
[9] Z. Shelby, K. Hartke, C. Bormann, and B. Frank. Constrained Application Protocol (CoAP). draft-ietf-core-coap-18. s.l. : IETF, 2013.
[10] S. Deering and R. Hinden. Internet Protocol, Version 6 (IPv6) Specification. RFC2460. s.l. : IETF, 1998.
[11] G. Montenegro, N. Kushalnagar, J. Hui and D. Culler. Transmission of
IPv6 Packets over IEEE 802.15.4 Networks. RFC4944: s.l. : IETF, Sep
2007.
[12] J. Hui and P. Thubert. Compression Format for IPv6 Datagrams over
IEEE 802.15.4-Based Networks. RFC6282. s.l. : IETF, 2011.
[13] IEEE Computer Society. IEEE Standard 802.15.4-2011. [Online].
Available: http://standards.ieee.org/getieee802/download/802.15.4-2011.
pdf
[14] IEEE Computer Society. IEEE 802.15 WPAN Task Group 4e (TG4e).
[Online]. Available: http://www.ieee802.org/15/pub/TG4e.html
[15] T. Winter, P. Thubert, A. Brandt, J. Hui, R. Kelsey, K. Pister, R. Struik,
JP. Vasseur, and R. Alexander. RPL: IPv6 Routing Protocol for LowPower and Lossy Networks. RFC6550. s.l. : IETF, 2012.