1. of processing. However, when they are coordinated




networks are dense of many small, inexpensive devices, called sensor nodes,
which are distributed to perform a specific action or an application-oriented
task. 1 Each sensor node can only do a limited amount of processing. However,
when they are coordinated with the information from a huge number of other
nodes, they are capable of measuring a physical environment in great details.
2 It can collect, process, analyse and disseminate data anytime anywhere.

the past few decades, traditional sensor network had been dominating in various
domains in term of data collection and functions monitoring.  However, when it requires thousands or
millions of sensor nodes spanning across a huge area, it will need miles of
shielded cable connections, which are costly and time consuming to build. 3
Hence, this motivates a huge effort in industrial investment and research
activities in Wireless Sensor Network (WSN) since the last decade. 4

WSN requires almost no infrastructure. 5 The exact location of sensor nodes
do not need to be predetermined or engineered. 6 It can be placed at harsh
and non-reachable environment such as mountains, over the sea and deep forest.
It is flexible as additional work station can be added at an ad-hoc condition.
The implementation cost is incredibly lower compared to traditional sensor
network. 7

WSNs are promising
approach for a variety of applications in biomedical health monitoring, wildlife tracking, natural disaster predicting, smart building, monitoring safety of buildings,
and industrial application. Detailed application will be further explored in
the review later on.

literature research aimed to provide an overview of sensor network. These are
the few main topics that will be covered and discussed:

To discuss the history and evolution of
sensor network
To understand the technology behind
sensor network
To discover the application of sensor
network in today’s world
To identify the limitation and security issue of sensor network

on the potential use of large sensor network on the monitoring
the behaviour of an engineering structure accurately during commissioning and
operation with a specific focus on vibration monitoring will be analysed in the discussion.

History & Evolution of
Sensor Network

            Similar to most of
the technologies, research and development on sensor network started with
military purpose. Sound
Surveillance System (SOSUS), a system of acoustic sensor on the ocean bottom,
was deployed to detect quiet Soviet submarines at strategic locations during
Cold War. More refined acoustic networks were developed over the next few years
for submarine surveillance. 8

research on Distributed Sensor Network (DSN) was carried out by the United States Defence Advanced Research Projects Agency
(DARPA) in early 1980s for US military. At that time, the Advanced Research
Projects Agency Network (ARPANET) had been in operation for several years, with
about 200 hosts at research institutes and universities. DSNs were assumed to
have a lot of spatially distributed low-cost sensing nodes, collaborating with
each other but operated autonomously, with information being routed to
whichever node that can best use the information. 8 The size of the sensors
were huge (approximately the size of shoes box or bigger) and thus the
applications were limited. Even though the researchers of DSN back then had the
vision of WSN in mind, but the technology was not quite ready yet. Hence, the
earliest DSNs were not tightly associated with wireless connectivity. 9

            Advancement of in communication,
microelectromechanical and computing technology in late 1990s have caused a new
wave of research in WSNs. The significant shift in WSNs research has attracted
large investment and international attention. Researches were focus on
developing sensor nodes that are cheaper, smaller, and suitable for highly
dynamic ad hoc and resource-constrained environments. Hence, many new civilian
applications of sensor networks such as vehicular sensor network, environment
monitoring and body sensor network have emerged. 10

            SensIT 11, an initiative research
program launched by DARPA has provided the current sensor networks with new
capabilities such as dynamic querying and tasking, ad hoc networking,
multi-tasking and reprogramming.

            Currently, WSNs have been believed
as one of the most important technologies in 21st century. 12 China has
included WSNs in their national strategic research programmes 13 which leads
to the acceleration and commercialization of WSNs as well as emerging of many
new technology companies such as Crossbow Technology and Dust Networks.

adoption of WSNs also drove the growth of Internet of Things (IoT). The integration
of cloud technologies and WSN technologies are the key elements of industrial
IoT. Both short range and long range of industrial IoT landscapes have emerged.

            According to ON WORLD, by 2021, there will be
33 million installed wireless devices used for industrial sensing and asset
tracking applications. Wireless tracking, sensing, equipment control and
associated services will reach $35 billion over the next five years for
industrial agriculture, construction, automation and related markets. 14

Figure 2.1: Global
Installed Industrial WSN Devices by Market Segment (2016-2021) 14


            Over the past few years, research on
Low Power
Wide Area Network (LPWAN) has been rolled out. It can communicate at a longer
range (up to 30 kilometres), lower cost and minimal maintenance. Hence, LPWAN potentially
will be another evolution from the sensor network technology. 14






Technology of Sensor Networks

3.1 Hardware Structure of a Sensor Node

A sensor node is made up of four basic components
such as sensing unit, processing unit, transceiver unit and a power unit which
is shown in Figure 3.1.

3.1: The component of a sensor node 15


Sensing units generally consists of 2 subunits: sensor and
analogue-to-digital converter (ADC). Based on the observed phenomenon, the
analogue signals produced by the sensors are converted to digital signals by
the ADC, and transmit into the processing unit. 6

            Processing unit is
usually associated with a small storage unit. It can manage the processes that
make the sensor node collaborate with the other sensor nodes to carry out the
assigned sensing tasks. 15 The
processing unit of a sensor node determines to most of the energy consumption and
computational capabilities of a sensor node. 16 The requirement of storage
size in term of fast and non-volatile memory can varies depending on the
overall sensor network structure. 17

unit is responsible of connection and communication between sensor nodes in a
network. There are a few choices of wireless transmission media such as Radio
Frequency (RF), Laser and Infrared. RF based communication is widely used
because it fits to most of WSN applications. 10

            Power unit is the most
critical component in a sensor node. Power is consumed by the other 3 units and
can be stored in batteries or capacitors for WSNs. Energy harvesting technique
is an alternative solution to solve the lifetime problem of a sensor node.
Energy can be harvested from the environment (solar, vibration, waves) or other
energy sources (body heat, foot strikes, fingers motion) and convert it to
electric energy to power the sensor node. 18

            Position finding system
may be required since most of the sensor networks require high location
accuracy and routing techniques.  Mobilizer
will be needed when it requires to move sensor nodes to execute assigned tasks.


3.2 Operating systems of Sensor

the years, there are various Operating Systems (OSes) emerging in the sensor
network community such as TinyOS, Contiki, SOS,
Mantis OS, Nano-RK, RETOS and LiteOS. 6 OS’s
role is to build reliable applications that are efficient and safe. 10 These
operating systems are increasingly mature and widely used in real-world
applications. Due to resource constrain in hardware platforms and dynamically
changing environment, development of WSN applications still remain as a
challenge. The choice of the operating system for WSN is critical to mitigating
these challenges. 19 Comparison will be done
on the 3 most commonly used sensor network OSes: TinyOS, Contiki and LiteOS.


TinyOS is a tiny component-based
operating system specifically designed for sensor network 10. This operating
system follows an event-driven programming model 21 and implemented in a special
programming language called NesC 22. The endorsement of NesC by TinyOS
consumes less resources and reduces development complexity. These features have
led to the widespread adoption of TinyOS in the WSN domain. Full application
must be replaced to perform software reconfiguration as static optimization
does not preserved component structure after compilation. 23


            Contiki OS is a light
weight and flexible operating for tiny network sensors that provide IP
communication. Contiki OS kernel is event-driven and use C as the programming
language which allows it to be highly portable. The system support
multithreading and Protothreads (a thread-like concurrency model). 24 Contiki achieved software
reconfiguration through a dynamic linking and loading module. 25


LiteOS is
multi-threaded operating system that provides Unix-like abstraction to wireless
sensor network. LiteOS support LiteC++ (an extension of C) and allows dynamic
loading, online debugging, and file system assisted communication stacks. It
supports software reconfiguration through a separation between kernel and user
application. 26       


Table 3.1: Comparison between TinyOS, Contiki and
LiteOS 20


3.3 Sensor Network Architecture

Figure 3.2: Sensor nodes scatter in a
sensor field. 6

sensor nodes are normally scattered in a sensor field as shown in Figure 3.2.
All the scattered sensor nodes in the sensor field are able to collect data and
route data back to the sink. Data are routed back to the sink by a multi-hop
infrastructure-less architecture through the sink as shown in Figure 3.2. The
sink is able to communicate with the task manager node through the Internet or
satellite. 6

Figure 3.3: The sensor network
protocol. 29

The protocol stack used by all sensor
nodes and the sink is shown in Figure 3.3. This protocol stack integrates data with networking protocols, promotes
cooperative efforts of sensor nodes, combines power and routing awareness, and communicates
power efficiently through the wireless medium. 5 The protocol stack
consists of 5 layers (application layer, transport layer, network
layer, data link layer, physical layer) and 3 planes (power management
plane, mobility management plane, task management plane).