smart cities

Article
Advanced Adaptive Street Lighting Systems for
Smart Cities
Gianfranco Gagliardi 1 , Marco Lupia 1 , Gianni Cario 1 , Francesco Tedesco 1, * ,
Francesco Cicchello Gaccio 2, , Fabrizio Lo Scudo 2 and Alessandro Casavola 1
1 Dipartimento di Ingegneria Elettronica, Informatica e Sistemistica (DIMES), Universitá della Calabria,
via Pietro Bucci 42c, 87036 Rende, Italy; [email protected] (G.G.); [email protected] (M.L.);
[email protected] (G.C.); [email protected] (A.C.)
2 Gavi.it S.r.l., Via Marina, 88812 Crucoli, Italy; [email protected] (F.C.G.); [email protected] (F.L.S.)
* Correspondence: [email protected]




Received: 6 November 2020; Accepted: 2 December 2020; Published: 7 December 2020


Abstract: This paper reports the results of a recently concluded R&D project, SCALS (Smart Cities
Adaptive Lighting System), which aimed at the development of all hardware/software components
of an adaptive urban smart lighting architecture allowing municipalities to manage and control public
street lighting lamps. The system is capable to autonomously adjust street lamps’ brightness on the
basis of the presence of vehicles (busses/trucks, cars, motorcycles and bikes) and/or pedestrians in
specific areas or segments of the streets/roads of interest to reduce the energy consumption. The main
contribution of this work is to design a low cost smart lighting system and, at same time, to define
an IoT infrastructure where each lighting pole is an element of a network that can increase their
amplitude. More generally, the proposed smart infrastructure can be viewed as the basis of a wider
technological architecture aimed at offering value-added services for sustainable cities. The smart
architecture combines various sub-systems (local controllers, motion sensors, video-cameras, weather
sensors) and electronic devices, each of them in charge of performing specific operations: remote
street segments lamp management, single street lamp brightness control, video processing for vehicles
motion detection and classification, wireless and wired data exchanges, power consumptions analysis
and traffic evaluation. Two pilot sites have been built up in the project where the smart architecture
has been tested and validated in real scenarios. Experimental results show that energy savings of up
to 80% are possible compared to a traditional street lamp system.

Keywords: smart lighting; smart cities; Internet of Things; video processing; lighting control; ZigBee
communication; energy saving

1. Introduction
According to [1,2], 80% of currently produced electrical power is used to supply urban necessities
and about 60% out of this is required for street lamps, due to their continuous operations at night time.
Thus, energy saving is a key point in the context of smart cities; in this paper, we focus on the achievable
energy savings by adopting smart lighting systems in place of traditional incandescent/fluorescent lamps.
In this respect, smart lighting control systems play a crucial role in energy consumption reduction.
In fact, the advances in wired and wireless networks, control technologies and embedded systems
have made it feasible to design modern lighting systems endowed with smart technologies where the
energy savings issue can be efficiently dealth with.
The Internet-of-Things (IoT) paradigm has emerged as a new network framework in which a large
variety of devices and objects are embedded with sensors, software and other technologies for the
purpose of connecting and exchanging data with other devices and systems over the Internet thanks

Smart Cities 2020, 3, 1495–1512; doi:10.3390/smartcities3040071 www.mdpi.com/journal/smartcities

Smart Cities 2020, 3 1496

to communication technologies such as WiFi, GSM, ZigBee, etc. [3]. The choice of communication
technology plays a key role in the design of an IoT architecture and is closely related to the application
into consideration. As an example, in applications such as wearable IoT devices where a body sensor
node is normally equipped with a very limited energy source, the life-cycle of the sensor node heavily
depends on the battery dimensions and on the processing and communication duty-cycling. For these
reasons, many research efforts are focusing on the use of low power communication technologies,
with the aim of optimizing energy consumption [4]. However, in order to arrive to a feasible and
cost-effective integration of different applications, it is necessary to find a compromise between size,
costs and connectivity range of the network devices. As an example, short range wireless technologies,
such as WiFi network devices, are not useful from an energetic point of view whilst long range
technologies, like GSM (Global System for Mobile Communications), are too expensive and are not
suitable for IoT applications. In recent years, Low-Power Wide Area Networks (LP-WANs) have
gained a lot of consideration in the IoT context due to their low power consumption and long range
capabilities. In this respect, long-range (LoRa) and narrow-band (NB) devices [5–7] are LPWAN
leading technologies in the design of Wireless Sensors Networks (WSNs) and are considered as the
basis for IoT applications (e.g., agriculture, medicine, industry, etc.). Smart cities are probably the
most suitable application context for WSNs allowing the readily development of smart micro-grid,
smart lighting, video surveillance and so on [8]. LoRa and ZigBee protocols have gained an important
role as wireless middle-wares due to their simplicity and their technical specifications that allows one
to enable power efficient wireless communication over long distances.
In recent years, smart lighting systems have received global attention and several research
programs have been carried out to show the potential and the positive impact provided by smart
technologies aimed at increasing the efficiency of street lamps management. These technologies embed
street lamps with sensors, control algorithms and wireless communications to build a lighting solutions
able to work autonomously in the Internet of Things (IoT) context [9].
In general, a Smart Street can be defined as a street made up of multifunctional smart lighting
poles equipped with various devices allowing them to communicate each other and share data with
a remote management system. The main goals of a smart street lighting system include a better
and more flexible exploitation of the urban environment, especially in terms of energy savings and
costs reduction.
Many studies and projects on smart lighting infrastructures have been presented in the literature
covering various aspects of a smart street lighting system. Among them, in [10] the technical aspects
of the design of smart infrastructures for public lighting networks based on the IEEE 802.15.4
short-range communication technology are investigated. A case study concerning a smart-building
application based on the LoRa low-rate, long-range communication protocol is also presented. In [11],
the advantages of using LED DC lamps with respect to traditional street lamps are emphasized. In fact,
the LED technology is characterized by longer lifetime, lower maintenance costs, higher efficiency,
and, most of all, the lamp are mercury-free and turn out to be easily disposable. In [12] an overview
about the smart lighting system for outdoor and indoor applications is presented whilst in [13] the
impact of LED street lighting is analyzed and changes of the street lighting municipal regulations are
proposed to promote and ensure the adoption of the LED technology. In [14], a prototype of smart
street-lighting system is presented, where a number of DC street lights are powered by photovoltaic
(PV) panels. The prototype includes a motion sensing circuit and exploits a battery to store the excess
energy harvested by the solar panel. The overall result is a smart and efficient street lighting system.
Further energy consumption reduction can be achieved if the LED lamps intensity is controlled
by a logic able to compute, on the basis of current traffic and weather information, suitable brightness
set-points for specific areas or segments of the street so as to implement a sort of “on-demand” lighting
system. The reference [15] presents a case study of an adaptive street lighting system based on a
model that, from a theoretical point of view, can ensure energy saving up to 88% with respect to using
conventional sodium lamps.
Smart Cities 2020, 3 1497

Here, the main idea is to maintain the lamps’ brightness at their minimum allowable level in a
certain segment of the street that is compatible with national/regional standards and/or safety limits
for both vehicular and pedestrian traffic. These aspects have been widely dealt with in many studies
where various, from simple to more elaborated, brightness reductions strategies have been proposed.
These strategies can be very sophisticated and can modulate the minimum allowable brightness on
different levels depending on if a single or many vehicles are present in the street segment of interest,
while also taking into account the presence of bikers or motorcyclists along with cars or trucks [16].
Further aspects to be considered in setting down the lamp brightness are related to citizen security
and pedestrian and driver safety. The lamp brightness should be so that the minimum allowable
level is compliant with the security standard. In view of this, smart lighting and sensors (e.g., camera
endowed with video analysis electronic device) embedded in street lights may reduce local disturbance
by alerting law enforcement to potential problems by capturing anomalous behavior. Furthermore,
the pedestrians need enough brightness in walking along the pavements and crossing the streets in
order to avoid obstacles and unevenness. In the same way, the drivers need to see obstacles within
their safety braking distance for comfortable and safe driving, which is a direct function of the vehicle
speed and current weather conditions.
Ways to implement these concepts have been discussed, e.g., in [17,18], where a specific
wireless network based on the ZigBee protocol (Wi-Fi, based on IEEE 802.11, is a very efficient
wireless technology; however, it is optimized for large data transfers using high-speed throughput,
rather than low power consumption. As such, Wi-Fi is unsuitable for low-power (coin cell)
operation. On the other hand, ZigBee technology is designed to carry small amounts of data
over a short distance while consuming very little power. As opposed to WiFi, it is a mesh
networking standard, meaning each node in the network is connected to each other. When speaking
specifically about power consumption, ZigBee-based networks generally consume 25% of the
power of WiFi networks. (https://www.digikey.com/en/articles/comparing-low-power-wireless-
technologies—https://www.link-labs.com/blog/zigbee-vs-wifi-802-11ah) has been presented to
attend the necessity of a distributed smart street lighting system, capable to control and monitoring
many areas containing a large number of street lamps. Similar approaches have been considered
in [19,20], where a LED system equipped with sensors and control technologies is used to undertake
vehicles motion detection to regulate the brightness of the lamps.
In this paper, the outcomes of a recent project (SCALS—Smart Cities Adaptive Lighting System)
focused on the design and the experimentation of a smart lighting control are presented. The system is
based on the use of local sensors, local controllers, video-processing algorithms for motion detection
and remote web application. The main contribution of this work is to design a low cost lighting
system capable to regulate the street light intensity on the basis of traffic presence in a monitored
street. At same time, the work aims to design an IoT infrastructure where each lighting pole is an
element of a network that can increase continuously their amplitude. More generally, the proposed
smart infrastructure can be viewed as the basis of a wider technological architecture aimed at offering
services (e.g., energy management, environmental monitoring, traffic monitoring and management,
decision process support, etc.) for sustainable cities. In this context, a lighting pole plays a key role
and can be viewed as an enabling element for value-added services. In this respect, the main idea is
to endow street lamps with electronic devices designed and customized to control the light intensity
as a function of vehicles and pedestrian presence in the accounted area. Furthermore, street lamps
are organized as a WSN that, through ZigBee communication, can share data among them and with
a network coordinator. The experimentations have shown that the proposed architecture is able to
provide the following advantages:

• energy savings;
• taxes reduction for the community;
• improving general well-being.
Smart Cities 2020, 3 1498

Another important issue relies on the idea that adopting a smart lighting systems able to adapt
automatically in real-time the light intensity and its features (i.e., luminance, flux, etc.) as a function of
“street actual needs”, is nowadays considered as a key instrument to reduce energy waste, light pollution
and, in a more broader view, greenhouse gas emissions [21–23].
In this respect, this paper describes all details about the implementation of the smart lighting
architecture and presents a significant added value with respect to the preliminary results of
this work published in [24]. The design choices concerning the infrastructure working modes,
network architecture, remote management web application and video processing procedure are
presented here in some details. In particular, it can be highlighted that the network architecture
exploits the ZigBee communication protocol that allows one to overcome the limitations due to the
use of power line communication (PLC) protocols, used in preliminary studies. In fact, unlike the
PLC technology whose efficiency depends on the state of the power line, the ZigBee protocol allows
one to establish a long distance multi-point secure network connection regardless of the state of the
transmission medium.
Furthermore, the video processing algorithm used here for traffic detection and classification is
described in full detail. The details of the protection and integrity of the data exchanged with the
remote web application are also accounted. In this respect, in addition to access control based on the
user credentials, the functionalities of the HyperText Transfer Protocol Secure (HTTPS) are exploited.
The rest of the paper is organized as follows: Section 2 presents the main features of the proposed
smart lighting system. The network infrastructure is described in Section 3 while Section 4 is devoted
to illustrate the single lamp lighting controller and its interactions with the other electronic devices.
Section 5 briefly describes the remote web application while in Section 6 the video processing
algorithms used for vehicles motion detection is described. Finally, Section 7 reports the results
obtained in two pilot sites and some conclusions end the paper.

2. Adaptive Smart Street Lighting Architecture

The proposed system is focused on the definition of an intelligent public street lighting
infrastructure that can ensure:

• the remote management of the street lighting system;

• an intelligent control of the street lighting lamps by providing the necessary light intensity at each
lamp/group-of-lamps as a function of daily time, vehicle and/or pedestrian presence and/or
weather conditions.

Figure 1 depicts the infrastructure concept which provide the services and facilities listed below:

1. Single or group of street lamps control;

2. A network of local communication devices based on the IEEE 802.15.4 specifications and realized
by the high-level Zigbee communication protocol;
3. Traffic and emergency conditions (e.g., road congestions and/or accidents) monitoring;
4. Consumptions evaluation;
5. Remote control and access to all functionalities for configuring the daily lighting schedules and
monitoring alarms through web application;
6. Remote communication, between smart-phones and the web application, based on 3G/4G/
WiFi connections.
Smart Cities 2020, 3 1499

Remote Management Station Smart pole n.4

- Traffic data
- Meteo data
- Traffic data - Emergency conditions
- Meteo data
Smart pole n.3 - Lighting control signal
- Emergency
- Remote commands
Smart pole n.2

Smart pole n.1

Gateway

3G/4G/Wi-Fi Communication
ZigBee Communication

Figure 1. Smart Street Lighting Concept

The smart infrastructure highlights the presence of the so-called smart pole. A smart pole is a
street light lamp mainly composed by a local control device, a video camera and a computational
unit for video processing, weather sensors, communication devices and highly efficient LED lamps.
The smart poles can exchange data (control commands and information requests) among them and,
through a gateway, with a remote management web application. The smart-pole working modes can
be summarized as follows:
• day time—Traffic monitoring functions are only enabled: all traffic data are extracted in real-time
through video processing and shared with the web application.
• night time— All smart functionalities are enabled and single/group LED lamp dimming is
performed on the basis of vehicles or pedestrians presence in the monitored area (street segment,
roundabout, square, etc...). In particular, if no vehicular/pedestrian presence is detected in the
monitored area, the LED light intensity is kept fixed to the minimum value prescribed by the
street lighting regulations. Otherwise, if a presence is detected, the LED light intensity is increased
up to the most suitable value for the monitored area, optionally based also on the number and
kind of vehicles detected and on the current weather conditions. In such a way, the LED light
intensity is considered as a function of vehicular/pedestrian presence and, in turn, a huge energy
saving can be achieved whenever the traffic is null for long periods at night time.
• failsafe default emergency condition—If, due to a loss of connectivity or to a component damage,
the smart lighting system fails a default condition is activated in order to avoid that the general
safety may be compromised. In this condition, all smart functionalities are disabled and the
system works as a traditional unconnected lighting system.
In view of this, it is noticeable that smartness of the infrastructure is achieved by the following
main steps:
1. Detection of vehicle and pedestrian presence in the monitored area;
2. Lamp brightness modulation as a function of daily hours, presence and weather conditions
(ambient brightness, rain, fog, etc.);
3. Providing, in all operating conditions, enough level of brightness as provided by national/regional
standard rules;
4. Sharing information and alerting the remote management and control station.
It is worth noticing that in some scenarios there is no need to equip all poles with motion detection
devices. For instance, in a straight road without side access, the smart lighting system can be configured
Smart Cities 2020, 3 1500

so that only two poles (at the beginning and at the end of the road) are equipped with motion detection
devices. In such a way two kind of smart poles can be defined:

• smart master pole, equipped with motion detection, weather station and dimming devices;
• smart slave pole, equipped with dimming devices only.

On the other hand, it should be highlighted that in some specific scenario (e.g., square, crossroads,
etc.), although the smart lighting system is able to verify in real-time both vehicular and pedestrian
traffic and then provide in all conditions, the right lighting level might not be necessary to equip the
lighting pole with motion detection device. In such a condition, the smart system can be equipped
only with a dimming device and remotely set to work so that the necessary light intensity at each
lamp/group-of-lamps is provided as a function of daily time.
Figure 2 shows the flow charts representing the operating procedure of master (a) and slave
(b) poles. In detail, after the initialization step, following the operating procedure the master pole
holds the current lighting setting until a change in either the ambient brightness or the weather
or traffic conditions is detected. In such case, according with the new conditions, the master pole
sets a new lighting level and generates an external event that is shared with the slave poles via the
local communication channel. On the other hand, the slave pole holds the current lighting setting
until an external event is received. In this case, the lighting level is updated according to the master
pole command.

M0
Board Initialization
S0
Board Initialization
M1
Running

S1
Ambient Yes M2
Running
Brightness Lighting Level Setting
Variation? and
External Event Generation

No
No S2
Lighting Level Setting
Weather Yes
Condition
Variation?

No No Yes
External
Event?

No Traffic
Yes
Variation?

(a) (b)

Figure 2. Flow charts: master pole (a) master and slave pole (b).

3. Network Infrastructure
The system architecture (Figure 3) exploits local and remote communication channels. The local
communication channel allows data exchange amongst the smart poles. On the other hand, the remote
communication channel enables the data exchange between the smart poles and the remote server
and, consequently, allows the centralized consultation of the collected data (i.e., power consumption,
traffic condition in different streets, weather conditions and so on).
The communication amongst the smart poles is based on the 802.15.4 IEEE ZigBee wireless
protocol, that is an open and global standard for communication among wireless devices. One of the
main advantages in the use of Zigbee devices is that they can easily be integrated into IoT applications
via Zigbee gateways. Another important aspect in the use of such kind of protocol is that the 802.15.4
IEEE standard provides the ZigBee protocol stack MAC layer that gives basic security services and
interoperability between devices. The basic security services allows to maintain an access control list
Smart Cities 2020, 3 1501

and the use of symmetric encryption algorithm to protect the transmission of data [25]. In a ZigBee
network, the devices can be configured as:

• coordinator/gateway: in this configuration the device is responsible for forming the network and
routing traffic to other devices. Every network has to have only one coordinator.
• router: the device can route traffic to other devices.
• end device: the device can not route traffic to other devices.

As depicted in Figure 3, the system architecture provides that:

• multiple local network can be configured;

• each local network is managed trough a coordinator (gateway) device;
• the electronic devices (e.g., video processing board, lighting control board and all other devices
installed in the system) belonging to each local network are equipped with a ZigBee radio module;
• the radio nodes acts as router and are organized in a mesh topology where each node connects to
the signal and passes it on to the next device.

From a theoretical point of view, the Zigbee protocol can handle up to 65,000 devices in a network
(https://zigbeealliance.org/solution/zigbee/). However in real application, due to the presence of
environment disturbance elements (e.g., obstacles as building and trees or other transmission devices)
the network size has to be strongly reduced in order to maintain, at the same time, signal strength and
efficiency in sending and receiving data. As a consequence, the network size is a parameter that can vary
on the basis of both application and environmental contexts. Therefore, to efficiently size the network, it is
recommended the use of specific software (e.g., XCTU configurations platform for XBee Solutions from
DiGi International Inc., (Hopkins, MN, USA) https://www.digi.com/products/embedded-systems/
digi-xbee/digi-xbee-tools/xctu) that allows to check the quality of the link between the connected
devices, the device addresses and other network parameters.

L.C.B. Lighting Control Board

L.C.B. S.C. Smart Camera
S.C. Server Web
S.W.
Gateway
L.C.B. Smart Lighting Pole
S.C.
L.C.B. ZigBee Communication
S.C.

L.C.B.
S.C.
S.W.
L.C.B.
S.C.
L.C.B. 3G / 4G / WiFi
S.C.

L.C.B.
S.C. L.C.B.
S.C.

Users
L.C.B.
S.C.
L.C.B.
S.C. Sensors Weather Stations

Figure 3. System architecture.

The network exploits various routing policies for the exchange of messages (command or
information requests) between smart poles. In such way, as shown in Figure 4, the gateway can
send information-requests/commands to a single smart pole (unicast routing), a group of smart poles
(multicast routing) or all smart poles connected to the network (broadcast routing).
The communication between the network coordinator (gateway) and the server or the users’
smart-phones is performed through a 3G/4G/WiFi communication network.
Smart Cities 2020, 3 1502

Furthermore, it is highlighted that the communication between the network nodes, the gateway
and the smart-phones exploits a custom communication whose details can be found in [24].

Gateway Gateway Gateway

L.C.B. L.C.B.
L.C.B.
S.C. S.C.
S.C.
L.C.B. L.C.B.
L.C.B.
S.C. S.C.
S.C.
L.C.B. L.C.B. L.C.B.
L.C.B. L.C.B. L.C.B.
S.C. S.C. S.C.
S.C. S.C. S.C.
L.C.B. L.C.B. L.C.B.
S.C. S.C. S.C.
(a) (b) (c)

Figure 4. Message routing: unicast (a), multicast (b) and broadcast (c).

4. Lighting Control System

The lighting control system is in charge to adjust the lamps brightness in real-time on the basis
of daily time and traffic intensity and classification so that the power consumption is minimized
and proper lighting levels are always guaranteed in every traffic and weather condition. The control
system mainly exploits two electronic devices (Smart Camera and Lighting Control Board) and a
network link between them. Figure 5 shows the lighting control procedure and the basic idea in the
interaction between the smart camera and the lighting control board. Both devices have been designed,
assembled and properly customized for our smart lighting application and equipped with a ZigBee
device in order to exchange data (information request/command) with the other smart cameras and
control boards. The communication between the smart camera and lighting control boards exploits the
communication protocol previously described.

Smart Camera Lighting Control Board

Video Frame

Video
Processing
No

EVENT
No received?

Motion ZigBee Yes

detected? Communication
Generate and send
dimming command
Yes

Generate and send

“motion detected”
EVENT

Figure 5. Lighting control procedure and boards interaction.

The Smart Camera can process video streaming. It implements the video processing algorithm
reported in Section 6 to detect moving objects in the monitored area. Hence, when either vehicles
or pedestrians are detected a ‘motion detected’ event is generated by the master pole who detected
the event and sent, via the local communication channel (Zigbee based communication), to all poles
equipped with the lighting control board.
On the other hand, the control board, complying with the International Electrotechnical
Commission Standards (Standard 60929 Annex E), allows one to adjust the intensity level by adopting
a 0–10 V analog lighting control protocol [26]. Basically, a control command applies a voltage between
Smart Cities 2020, 3 1503

0 and 10 DC volts to produce a varying light intensity level. The light output changes are proportional
to the DC voltage variations. The standard requires that the ballast (or driver) provides full light
output when the control voltage is 10 Volts. At a control voltage of 1 volt, the ballast (driver) provides
it’s minimum light level. Any voltage lower than 1 volt is defined as the minimum.
When the ‘motion detection’ event is received, the lighting control board generates a control
command that is conveyed to all poles present in the street segment under control. After a settable
number of seconds, if no further ‘motion detection’ events occur, the control board drives the LED
lamp to the minimum lighting values prescribed by the street lighting regulations.
In addition to the described control strategy, the lighting control board allows different control
modes. As an example, it can operate both by following daily lighting profiles (the light intensity of
the lamp is follows a temporal profile) or event-based scheduled values of light intensity. Furthermore,
the control architecture allows one to arrange the poles in different groups which can operate with
different control modalities.
The lighting control board is also in charge of:

• adjusting the lighting level as a function of environment conditions by exploiting data from
ambient brightness sensors and the weather station;
• detecting anomalous behaviors in the network of poles and faults on street lamps. If such
anomalous conditions occur, the board generates an event that is transmitted to the remote
web application and the failsafe default emergency operating mode is enabled.

5. Remote Management and Control via Web Application and Mobile APP
A further important aspect in a smart lighting systems is the friendliness level of interactions
between the system managers and the infrastructure. In view of this, our system is endowed with
a web application and mobile APP that allows remote management and control by providing the
following services:

• Account management that allows the manager to define the permissions and the access levels to
the web application and mobile APP (i.e., super-administrator, system manager, system operator,
user, etc.);
• Control modality selection;
• Lighting system monitoring and control;
• Anomalous conditions detection and alarms generation;
• Traffic and meteo information monitoring;
• Energy consumption evaluation.

The remote web application provides access to the smart lighting system settings. For instance
it is possible to set the lighting profile to each single street lamp or to a group of them. It is, also,
possible to set lighting level variation range so that abrupt change of brightness can be avoided.
Moreover, the user can evaluate information related to the current/averaged consumption and energy
saving, traffic intensity and classification, weather conditions and other alarms.
The remote web application, in addition to the typical access control based on user credentials
checking, exploits the Hypertext Transfer Protocol Secure (HTTPS) to improve the protection of the
privacy and the integrity of the exchanged data while in transit. Hypertext Transfer Protocol Secure
(HTTPS) is a variant of the standard web transfer protocol (HTTP) that adds a layer of security on the
data in transit through a secure socket layer (SSL) or transport layer security (TLS) protocol connection.
Thanks to the HTTPS properties, the web application enables encrypted communication and secure
connection between the remote users and the web server.
Figure 6 shows some illustrative pages of the web platform. In particular the street lighting
remote management page is depicted Figure 6a where the user can operate in order to set the operating
modes and configure the street poles operations. Moreover, information about street pole locations
Smart Cities 2020, 3 1504

(via GPS data displayed on street maps) and status, lighting system consumption trends and fault
alarms are displayed in this page. On the other hand, in Figure 6b is reported the pole management:
this page shows information related to consumption trends, GPS information, lighting profile, energy
costs evaluation and the energy saving granted by the smart system with respect to a traditional
(e.g., classic sodium vapor street lights) or LED street lamps system. Finally, Figure 6c shows the traffic
monitoring web page.

Connectivity Status

Street Lighting System

Configurations Menu

Street Lamps
GPS Locations

(b)
Alarms

Consumptions Trend (Last 24h)

Street Lamps
Informations

(a) (c)

Figure 6. Web application: remote management page (a), pole management page (b) and traffic
monitoring (c).

The mobile APP includes all functionalities provided by the web application and, at same time,
allows authorized users to manage the lighting levels of specific lamps. Thanks to this option, in the
event of street emergency, any user with granted access (i.e., a police officer) can interact with the
system and take the appropriate actions in order to improve the overall visibility.

6. Video Processing for Vehicles and Pedestrians Detection

Detection of moving objects (vehicles and pedestrians) is of paramount importance in traffic
detection and classification and it is also relevant in smart street lighting control applications when
traffic data are considered. Here, the aim of using video processing units is at detecting the presence of
moving vehicles or pedestrians in a video sequence. To this end, every frame of the video sequence is
split into two different logical areas [27]: the foreground, that collects pixels belonging to the objects of
interest (e.g., pedestrians, cars, bikes, etc.); the background, that groups all pixels belonging to objects
that remains fixed in all successive frames (e.g., buildings, sky, trees, etc.). The common approach
for discriminating moving object from the background scene is the background subtraction method.
The idea is to “subtract” the current image represented by a matrix Fi from a reference image Bi which
is acquired from a static background during a period of time. If such a subtraction is greater then a
certain threshold Th , i.e.,
| Fi − Bi | > Th (1)

then a moving objects is detected (Figure 7).

Several background subtraction algorithms have been proposed [27] and most of them are
based on the following four steps (Figure 7): image pre-processing, background modeling process,
foreground detection and binary mask creation process and binary image analysis.
Smart Cities 2020, 3 1505

Current Frame
Threshold
Foreground &
binary mask

- > Binary Image

Analysis

Detection of
moving objects
Background model

Figure 7. Background subtraction and detection processes: general approach.

The first step is an image processing task, in charge of converting the raw input video into a
format that can be easily processed by the successive steps. Typically during this task, the RGB
(Red-Green-Blue) image is converted into a gray-scale one. The second task is the background
modeling process that uses the first frame or previous frames to build up a background model [28].
The foreground detection and binary mask creation process compares the input video frame with the
background model and identifies candidate foreground pixels from the input frame [29]. A threshold
logic is then used to convert the gray scale image into a binary mask, so that the objects of interest are
separated from the background.
The last step is a binary image analysis finalized to eliminate the pixels that do not correspond to
the current moving object. Furthermore, the binary image is processed in order to detect the moving
objects and to identify vehicles and pedestrians.
Although many choices are possible, the adopted video-processing procedure exploits a
background modeling procedure based on the recursive mixture of Gaussians model described
in [30,31]. Besides the moving objects detection process, in order to classify vehicles and pedestrians,
the algebraic procedure based on the eigenface approach initially proposed in [32] for faces recognition,
here specialized for pedestrians and vehicles classification, is considered. This algorithm ensures an
efficient execution on low cost boards. Classification can be achieved by comparing how objects are
represented by the basis set composed by vehicles and pedestrians images (Figure 8).

(a) (b)

(c) (d)
Figure 8. Example of basis set for moving object classification: pedestrians (a) and vehicles (cars (b),
trucks (c) and bikes (d)).
Smart Cities 2020, 3 1506

However, it is worth noting that this classification procedure, although it is relatively fast and
accurate, is not immune from mis-classification episodes: target objects might have some spurious
feature that brings the algorithm to classify them in the wrong classes. For example, SUVs can easily
misunderstood as little trucks, pedestrians carrying voluminous objects cannot be detected as person,
bicycles with bags as motorcycles, and so on. Table 1 shows the error rates relative to the analysis
conducted over our hand-curated dataset composed by 100 elements for each kind of moving objects z.

Table 1. Classification procedure—Performance Analysis.

% Mis-Classification
Object % Success Pedest. Car Truck Bike
Pedestrian 93 - 0 0 7
Car 89 0 - 7 4
Truck 94 0 6 - 0
Bike 91 5 4 0 -

Furthermore, the procedure here detailed improves the preliminary one described in [24] where
the classification algorithm was based on a raw analysis of moving object dimensions.

7. Experimental Results
In order to demonstrate the effectiveness of the proposed smart street lighting system two
experimental pilot sites were built up. This choice allowed us to set up two different application
scenarios, which respectively refers to a busy road (experimentation site n.1) and to a secondary road
(experimentation site n.2). The first site is a road owned by the University of Calabria (Italy) whilst
the second one is owned by the municipal district of Crucoli (Italy).
The experimentation activity has been spitted in three different steps. The first step has concerned
the collection of data related to consumptions of the traditional lighting systems originally present in
the two demonstration sites.
Hence, these lighting systems consists of the following street lamps:

• Experimentation Site n.1: ten High Pressure Sodium (HPS) 250 [W] lamps with light intensity of
33.000 [lm];
• Experimentation Site n.2: ten High Pressure Sodium (HPS) 150 [W] lamps with light intensity of
17.700 [lm].

In the second step, for both experimentation sites, the existing HPS street lamps have been
upgraded to LED technology. After the upgrading, a new data collection campaign aimed at comparing
the consumptions before and after the LED lamps substitution, has been undertaken.
The last step has been focused on the installation of the smart devices in both the demonstration
sites and collecting again the consumption data. Figure 9a,b and Figure 10a–c show, respectively,
the state of the art experimentation site 1 and 2 before and after the upgrading works. In details,
Figure 9a shows the experimentation site 1 in standard configuration where 250 [W] High Pressure
Sodium lamps were present whilst Figure 9b shows the experimentation site in the Smart configuration.
On the other hand, Figure 10a shows the traditional lighting poles, 250 [W] High Pressure Sodium
lamps, that were originally used in the experimentation site 2 while Figure 10b,c shows the upgraded
smart lighting pole and the experimentation site in the Smart configuration.
Table 2 reports data (number, type, power and light intensity of lamps) about the lighting system
after the upgrading activities for both sites. It is important to note that the last two columns of Table 2
report the maximum and minimum light intensity, along with relative lamp powers, that are set-up
when traffic is or not detected.
Smart Cities 2020, 3 1507

Smart Camera
&
Lighting Control Board

Lighting Control Board

Smart Camera
&
Lighting Control Board

(a) (b)

Figure 9. Experimentation site n.1: standard (a) and smart lighting system (b) configurations.

Smart Camera
&
Lighting Contrl Board

Lighting Control Board

(a) (b) (c)

Figure 10. Experimentation site n.2: standard pole (a), smart pole (b) and smart lighting system
(c) configurations.

Table 2. Experimentation Sites—Smart System.

Site # Lamps Power Max Lamp Light Intensity Min Lamp Light Intensity
#1 10 135 [W] 12,500 [lm] at 135 [W] 3500 [lm] at 48 [W]
#2 90 [W] 8500 [lm] at 90 [W] 2500 [lm] at 26 [W]

The experimental activity, started in October 2018 and still ongoing, has been useful to compare
the consumption of our smart lighting system with the respect to both traditional and LED-based
systems. Figure 11 shows consumption comparison among the compared technologies for both the
considered demonstration sites. Hence, the upper sub-figures of Figure 11a,b report the instant power
trend from 17:10 h to 07:10 h whilst the lower sub-figures show the ”motion detected” event trend.
From these Figures it can be observed that when the ”motion detected” event (e0,1 ) is OFF the instant
power of the Smart System is at minimum value (≈480 [W] for site n.1 and ≈320 [W] for site n.2).
On the other hand, if a vehicle or a pedestrian is detected (i.e., the ”motion detected” event is ON) the
instant power of the Smart System is at maximum value (≈1500 [W] for site n.1 and ≈1080 [W] for site
n.2). It is important to note that these figures report, for a better view, data only from 17:10 h to 07:10 h.
Smart Cities 2020, 3 1508

3000 2000

2500
1500
2000

1500 1000 I
I

1000
500
500

0 0
17:10 19:10 21:10 23:10 01:10 03:10 05:10 07:10 17:10 19:10 21:10 23:10 01:10 03:10 05:10 07:10

ON ON

II
II

OFF OFF
17:10 19:10 21:10 23:10 01:10 03:10 05:10 07:10 17:10 19:10 21:10 23:10 01:10 03:10 05:10 07:10

(a) Site n.1 (b) Site n.2

Figure 11. Instant power trend [W]: busy (a) and secondary road (b).

Figure 12 shows the evaluation of the average daily consumption computed accounting the data
measured during the experimentation activity. The average daily consumption values are reported in
Table 3. Table 4 reports a comparisons of achieved energy saving. In particular, this table compares
the energy saving achieved by the proposed smart lighting system w.r.t. the standard lighting system
(column Smart vs. Std) and the LED lighting system (column Smart vs. LED). Furthermore the table
compares the energy saving capability of the LED lighting system w.r.t. the standard lighting system
(column LED vs. Std).

Table 3. Average daily consumption.

Site Std Lighting LED Lighting Smart Lighting

#1 37.14 [kWh] 21.51 [kWh] 11.01 [kWh]
#2 26.57 [kWh] 15.39 [kWh] 4.52 [kWh]

Table 4. Energy saving.

Site Smart vs. Std Smart vs. LED LED vs. Std
#1 70.35% 48.81% 42.07%
#2 82.99% 70.65% 42.07%

40 30

35
25

30

20
25

20 15

15
10

10

5
5

0 0
Smart LED Standard Smart LED Standard

(a) Site n.1 (b) Site n.2

Figure 12. Evaluation of average daily consumption [kWh]: busy (a) and secondary road (b).

It can be observed that the use of Smart technologies can ensure remarkably larger energy saving
with respect to the standard lighting systems and LED systems.
Smart Cities 2020, 3 1509

With reference to the consumptions reported in Table 3 and to the energy unit cost for Site n.1
(0.25 [e/kWh]) and Site n.2 (0.22 [e/kWh]), Table 5 reports a comparison of the energy annual costs
for both the experimentation sites:

annual cost = average daily consumption × days o f the year × energy unit cost

This table highlights that the smart lighting systems allows a remarkable saving.

Table 5. Energy Annual Costs.

Site Std Lighting LED Lighting Smart Lighting

#1 3390 e 1963 e 1005 e
#2 2133 e 1236 e 363 e

The experimentation activities are also useful to validate the traffic (vehicles/pedestrians)
detection and classification performances of smart lighting system. As stated in Section 2,
the smart-pole operates in day time and night time working modes. When the smart-pole works
in day time configuration all traffic data are extracted in real-time through a video processing procedure
and shared with the remote web application. On the other hand, when the smart-pole works in night
time single/group LED lamp dimming is performed on the basis of vehicles or pedestrians presence
in the monitored area. In Figure 13 is reported as an example of the output of the video processing
algorithm described in Section 6.

(a) (b)

Figure 13. Site n.1—day time (a) and night time (b) working modes.

Figure 14 reports the daily average traffic trend for both experimentation sites. The graph
reports the traffic detected during daylight hours (from 07:00 h to 18:00 h—day time working mode).
As expected, the traffic intensity in Site n.1 resulted in much larger traffic intensity than that in Site
n.2. Table 6 reports the averaged annual traffic intensity and the classification results of the moving
objects detected in both experimentation sites. The results show how the detection and classification
procedure ensures good performance because there is a low percentage of ‘false detection’ that the
system classifies as Unknown moving objects.
It is evident from the presented results that the proposed smart lighting technology can achieve
huge energy savings in comparison with traditional systems. In fact, energy savings are up to 82.99%
(while the energy savings pertaining to the LED substitution without smart functionality were limited
to 42.07%) have been achieved when compared to the traditional system and up to 70.65% with
respect to the LED upgraded system. Moreover, it is worth highlighting that the overall cost of the
investment in this type of smart lighting technology is low and, thanks to the reduction of energy costs,
is returned within two years from installation. Finally, the system is also capable of achieving good
traffic monitoring performance. Demonstration videos of the smart lighting system can be found at
the web link http://www.smartlightingproject.it/new/index.php/demo.
Smart Cities 2020, 3 1510

400 14

350 12

300
10

250
8

200
6
150

4
100

2
50

0 0
07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00

(a) Site n.1 (b) Site n.2

Figure 14. Daily average traffic trend: busy road (a) and secondary road (b).

Table 6. Experimentation Sites—Annual Average Traffic Intensity and Classification.

Site Objects Detected Car Truck Moto Pedestr. Unknown

#1 731,825 84.64% 0.74 9.03% 5.24% 0.35%
#2 52,560 26.4% 18.75% 37.5% 15.25% 2.10%

Remark to the Experimentation Activities and Future Application

The proposed system can be viewed as a valid solution for smart lighting applications. It is a
low-cost and easy implementable solution and can be customized to be useful in various application
contexts. It does not require extra maintenance activities and the achievable energy cost savings
justify the investment in this kind of technology. Finally, it is worth remarking that, thanks to its
characteristics, this smart lighting solution as been selected for the set-up of 180 lighting poles owned
by the municipal district of Capranica Prenestina—Rome (Italy).

8. Conclusions
In this paper, a smart street lighting prototype has been described and shown to be effective in
two experimental pilot sites. The proposed system has been designed so that it can be considered as
an IoT infrastructure where each lighting pole can be viewed as an enabling element of value-added
services (e.g., energy management, environmental monitoring and light pollution reduction, etc.)
for sustainable cities. From this point of view, the overall architecture is scalable, can increase its
functionality and can be implemented by connecting smart devices to the network. The paper first
introduced a general system description, then the basic ideas and the main features of the smart lighting
prototype have been presented. The system is able to establish the optimal lighting profile for each
lamps on the basis of both traffic and weather conditions. The experimental stage has demonstrated
that the proposed system is capable of achieving significant energy savings compared to traditional
street lamp systems and, at the same time, allows the performance of traffic classification with a low
rate of error. Furthermore, it is worthing pointing out that, thanks to the low implementation costs, the
investment in smart lighting systems may be recovered in a short period and this represents a further
advantage in the use of such technologies.

Author Contributions: Conceptualization, G.G.; Data curation, F.C.G.; Funding acquisition, G.G., M.L., G.C. and
A.C.; Investigation, F.L.S.; Methodology, G.G.; Project administration, G.G., M.L. and G.C.; Software, F.C.G. and
F.L.S.; Supervision, G.C. and A.C.; Writing—original draft, G.G. and F.T.; Writing—review & editing, F.T. and A.C.
All authors have read and agreed to the published version of the manuscript.
Funding: This research is based upon work partially supported by the R&D project J28C17000170006,
entitled ‘Smart Cities Adaptive Lighting System (SCALS)’, granted by the Calabria Region within the POR
CALABRIA FESR-FSE 2014-2020 Asse I, Obiettivo Specifico 1.2.
Smart Cities 2020, 3 1511

Conflicts of Interest: The authors declare no conflict of interest.

References
1. Palumbo, M.L. Architettura Produttiva: Principi di Progettazione Ecologica; Maggioli Editore: Milan, Italy, 2012;
ISBN 978-88-387-6849-8.
2. Jagadeesha, Y.M.; Akilesha, S.; Karthika, S. Prasantha, Intelligent Street Lights. Procedia Technol. 2015,
21, 547–551. [CrossRef]
3. Dudhe, P.V.; Kadam, N.V.; Hushangabade, R.M.; Deshmukh, M.S. Internet of Things (IOT): An overview and its
applications. In Proceedings of the 2017 International Conference on Energy, Communication, Data Analytics
and Soft Computing (ICECDS), Chennai, India, 1–2 August 2017.
4. Fortino, G.; Gravina, R.; Galzarano, S. Wearable Computing: From Modeling to Implementation of Wearable
Systems Based on Body Sensor Networks; John Wiley & Sons, Inc.: Hoboken, NJ, USA, 2018; ISBN 9781119078821.
5. Sarr, Y.; Gueye, B.; Sarr, C. Performance Analysis of a Smart Street Lighting application using LoRaWan.
In Proceedings of the 2019 International Conference on Advanced Communication Technologies and
Networking (CommNet), Rabat, Morocco, 12–14 April 2019. [CrossRef]
6. Sinha, R.S.; Seung-Hoon Hwang, Y.W. A survey on LPWA technology: LoRa and NB-IoT. ICT Express
2017, 3, 14–21. [CrossRef]
7. Zhu, H.; Chang, A.S.F.; Kalawsky, R.S.; Tsang, K.F.; Hancke, G.P.; Lo Bello, L.; Ling, W.K. Review of
State-of-the-Art Wireless Technologies and Applications in Smart Cities. In Proceedings of the IECON
2017—43rd Annual Conference of the IEEE Industrial Electronics Society, Beijing, China, 29 October–1 November
2017. [CrossRef]
8. Mehmood, Y.; Ahmad, F.; Yaqoob, I.; Adnane, A.; Imran, M.; Guizani, S. Internet-of-things-based smart cities:
Recent advances andchallenges. IEEE Commun. Mag. 2017, 55, 16–24. [CrossRef]
9. Castro, M.; Jara, A.J.; Skarmeta, A.F. Smart lighting solutions for smart cities. In Proceedings of the 27th IEEE
International Conference on Advanced Information Networking and Applications Workshops (WAINA),
Barcelona, Spain, 25–28 March 2013; pp. 1374–1379.
10. Pasolini, G.; Buratti, C.; Feltrin, L.; Zabini, F.; De Castro, C.; Verdone, R.; Andrisano, O. Smart City
Pilot Projects Using LoRa and IEEE802.15.4 Technologies. Sensors 2018, 18, 1118; doi:10.3390/s18041118.
[CrossRef] [PubMed]
11. Cheng, C.A.; Chang, C.H.; Ching, T.Y.; Yang, F.L. Design and implementation of a Single-Stage Driver
for supplying an LED Street-lighting Module with Power Factor Corrections. IEEE Trans. Power Electron.
2015, 30, 956–966. [CrossRef]
12. Barve, V. Smart Lighting for Smart Cities. In Proceedings of the 2017 IEEE Region 10 Symposium (TENSYMP),
Cochin, India, 14–16 July 2017.
13. Garcia, R.B.; Angulo, G.V.; Gonzalez, J.R.; Tavizon, E.F.; Cardozo, J.I.H. LED Street Lighting as a Strategy for
Climate Change Mitigation at Local Government Level. In Proceedings of the IEEE 2014 Global Humanitarian
Technology Conference, San Jose, CA, USA, 10–13 October 2014; pp. 345–349.
14. El-Faouri, F.S.; Sharaiha, M.; Bargouth, D.; Faza, A.Z. A smart street lighting system using solar energy.
In Proceedings of the 2016 IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT-Europe),
Ljubljana, Slovenia, 9–12 October 2016. [CrossRef]
15. Hannan, S.; Humayun Kabir, M.; Uddin, M.J. A Case Study on a Proposed Adaptive and Energy Efficient
Street Lighting System for Chittagong City. In Proceedings of the 1st International Conference on Advances
in Science, Engineering and Robotics Technology 2019 (ICASERT 2019), Dhaka, Bangladesh, 3–5 May 2019.
[CrossRef]
16. Mohamed, S.E. Smart Street Lighting Control and Monitoring System for Electrical Power Saving by Using
VANET. Int. J. Commun. Netw. Syst. Sci. 2013, 351–360, doi:10.4236/IJCNS.2013.68038. [CrossRef]
17. Denardin, G.W.; Barriquello, C.H.; Pinto, R.A.; Silva, M.F.; Campos, A.; do Prado, R.N. An intelligent system
for street lighting control and measurement. In Proceedings of the IEEE Industry Applications Society
Annual Meeting, Houston, TX, USA, 4–8 October 2009.
18. Denardin, G.W.; Barriquello, C.H.; Campos, A.; do Prado, R.N. An Intelligent System for Street Light
Monitoring and Control. In Proceedings of the Brazilian Power Electronics Conference, Bonito-Mato Grosso
do Sul, Brazil, 27 September–1 October 2009.
Smart Cities 2020, 3 1512

19. Vena, P.C.; Tharakan, P.; Haridas, H.; Ramya, K.; Joju, R.; Jyothis, T.S. Smart street light system based on
image processing. In Proceedings of the IEEE International Conference on Circuit, Power and Computing
Technologies (ICCPCT), Nagercoil, India, 18–19 March 2016.
20. Ikpehai, A.; Adebusi, B.; Kharel, R. Smart street lighting over narrowband PLC in a smart city: The Triangulum
case study. In Proceedings of the IEEE 21st International Workshop on Computer Aided Modelling and Design
of Communication Links and Networks, (CAMAD), Toronto, ON, Canada, 23–25 October 2016.
21. Sittoni, A.; Brunelli, D.; Macii, D.; Tosato, P.; Petri, D. Street lighting in smart cities: A simulation tool for
the design of systems based on narrowband PLC. In Proceedings of the 2015 IEEE First International Smart
Cities Conference (ISC2), Guadalajara, Mexico, 25–28 October 2015. [CrossRef]
22. Yusoff, Y.; Rosli, R.; Karnaluddin, M.; Samad, M. Towards smart street lighting system in Malaysia.
In Proceedings of the IEEE Symposium on Wireless Technology and Applications (ISWTA), Kuching,
Malaysia, 22–25 Sepember 2013; pp. 301–305.
23. Higuera, J.; Hertog, W.; Peralvarez, M.; Polo, J.; Carreras, J. Smart lighting system ISO/IEC/IEEE 21451
compatible. IEEE Sens. J. 2015, 15, 2595–2602. [CrossRef]
24. Gagliardi, G.; Casavola, A.; Lupia, M.; Cario, G.; Tedesco, F.; Lo Scudo, F.; Gaccio, F.C.; Augimeri, A. A smart
city adaptive lighting system. In Proceedings of the 3rd International Conference on Fog and Mobile Edge
Computing, FMEC 2018, Barcelona, Spain, 23–26 April 2018.
25. Fan, B. Analysis on the Security Architecture of ZigBee Based on IEEE 802.15.4. In Proceedings of the
IEEE 13th International Symposium on Autonomous Decentralized System (ISADS), Bangkok, Thailand,
22–24 March 2017.
26. Yan, L.; Chen, Y.; Chen, B. Integrated analog dimming controller for 0-10V dimming system. In Proceedings of
the 10th China International Forum on Solid State Lighting (ChinaSSL), Beijing, China, 10–12 November 2013.
27. Elhabian, S.Y.; El-Sayed, K.M.; Ahmed, S.H. Moving object detection in spatial domain using background
removal techniques-state-of-art. Recent Patents Comput. Sci. 2008, 1, 32–54. [CrossRef]
28. Xu, Y.; Dong, J.; Zhang, B.; Xu, D. Background modeling methods in video analysis: A review and
comparative evaluation. CAAI Trans. Intell. Technol. 2016, 1, 43–60. [CrossRef]
29. Zivkovic, Z.; Van Der Heijden, F. Efficient adaptive density estimation per image pixel for the task of
background subtraction. Pattern Recognit. Lett. 2006, 27, 773–780. [CrossRef]
30. Zivkovic, Z.; van der Heijden, F. Recursive Unsupervised Learning of Finite Mixture Models. IEEE Trans.
Pattern Anal. Mach. Intell. 2004, 26, 651–656. [CrossRef] [PubMed]
31. Zivkovic, Z. Improved adaptive Gaussian mixture model for background subtraction. In Proceedings of the
17th International Conference on Pattern Recognition, 2004. ICPR 2004, Cambridge, UK, 26 August 2004;
Volume 2, pp. 28–31.
32. Matthew, T.; Alex, P. Eigenfaces for recognition. J. Cogn. Neurosci. 1991, 3, 71–86.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).