Figure 1. Schematic of a Wireless Sensor Network
In the recent decades, sensors have affected on our life in various fields, including research activities, standardization procedures, and industrial investments. Various types of sensors (e.g., pressure, temperature, and humidity) can be used in Wireless Sensor Networks (WSNs) to design and implement some of the important applications, such as environmental monitoring, healthcare systems, and military issues. WSNs consist of the low-power, large-scale, and low-cost sensor nodes. The nodes measure phenomena data (e.g., smoke density) to transmit the measured data to a center (e.g., sink or base station). Routing, security, and fault tolerance are some of the crucial challenges in sensor networks. This paper describes various physical features and key usages of the popular sensors. Furthermore, three WSNs applications in monitoring, healthcare, and military are considered subsequently. Since sensor localization and data mining are two important topics in WSNs, their categories and characteristics are addressed too. Evaluation results show the performance of sensor networks compared to radio-frequency identification (RFID) in terms of utilization, purpose, components, mobility, power supply, programmability, and deployment. Besides, some of the WSNs simulators are compared to each other in terms of computation time and memory usage.
Nowadays, Wireless Sensor Networks (WSNs) include much more applications in industry, monitoring, medical, etc. They can be typically illustrated as a network of wireless sensor nodes, which monitor different environments (e.g., hospital) in a cooperated way. This goal can be achieved through an interaction between persons/computers and the environments in order to transmit the sensed data toward an assigned center (e.g., sink or base station). In the most cases, sensor nodes are randomly deployed in the desired environments without having any pre-defined structure. Sensor networks should involve the lowest hardware complexity and energy consumption to prolong the network lifetime) [8, 41]. They have various applications in research activities, standardization procedure, and industrial investments [15, 52]. Different types of sensors (e.g., pressure, temperature, and humidity) are applied in WSNs to monitor the desired environments [44]. Environmental monitoring, healthcare systems, and military usages are some of the most important applications which can be designed and implemented by sensor networks [5, 9, 21, 24, 49, 51].
Figure 1 shows schematic of a wireless sensor network. This network consists of the sensor nodes which are placed on buildings, cars, humans, etc. Each sensor measures phenomena data (e.g., temperature and humidity) in order to transmit the measured data to a center (e.g., a base station). This paper, firstly, presents a brief description of WSNs. It, secondly, explains various types of the physical sensors which are applied in WSNs applications. Thirdly, three popular WSNs applications are considered in the paper. Finally, the performance of sensor networks is evaluated under a simulation scenario compared to radiofrequency identification (RFID) [ 3, 4, 16, 28]. Moreover, some of the WSNs simulators are compared to each other.
Objectives of this study are categorized as (i) presenting main history and features of various sensors that can be used in WSNs, (ii) describing general information about three popular WSNs applications in monitoring, healthcare, and military, (iii) addressing different characteristics of the sensor localization and data mining, (iv) evaluating the performance of sensor networks compared to RFID networks, and (v) comparing some of the most used WSNs simulators to each other.
Figure 1. Schematic of a Wireless Sensor Network
The main contribution of this work is to describe various features of the popular physical sensors which are mainly used in industrial projects, explain different WSNs applications, and analyze comparison results of sensor networks under various simulation scenarios. According to the above goals, section 1 presents basic and essential features of the pressure, position, motion, temperature, humidity, biosensor, and chemical sensors. Afterward, section 3 describes main characteristics of sensor networks and three popular WSNs applications, including environmental monitoring, healthcare systems, and military usages. Section 4 explains several methods of sensor localization and data mining in sensor networks. Section 5 presents the comparison results of WSNs and RFID systems as well as the evaluation results of some WSNs simulators. Finally, the paper is concluded.
A sensor is an electronic device that measures phenomena data (e.g., temperature). It transmits the sensed data to a center (e.g., a monitoring system). Since the sensing process is one of the basic operations in sensor networks, sensors play an essential role in these networks. They have different features, such as sensitivity, accuracy, and dynamic range [50]. This section presents basic and essential features of the pressure, position, motion, temperature, humidity, biosensor, and chemical sensors.
Pressure sensor is one of the mechanical sensors, which measures the defined motion actions. Pressure microsensor is the first microsensor offered by industry. Piezoresistive, capacitive, and optical are the most popular pressure sensors that have been used in different applications. Piezo-resistive pressure sensor is one of the initial products of MEMS technology. It can be used in biomedical applications, household appliances, and automotive industry. A piezo-resistive pressure sensor has a piezoresistive effect that is integrated in a membrane. Capacitive sensor uses an alternating voltage and causes to the charges reverse their positions. This process causes to the alternative electric current which can be detected by a capacitive sensor. A membrane or comb is utilized in capacitive sensors on their surface to generate a membrane for the deflection action as well as a capacitance to can be changed by the sensor. Optical pressure sensor performs based on the principle of the Mach-Zehnder interferometer. Laser light is conducted into the sensor through an optical fiber. It is divided to two beams in a way that one beam crosses via another beam which is deformed by the pressure [13, 30]. Figure 2 shows several types of the pressure sensor.
Figure 2. Three types of the Pressure Sensor, (a) Piezo-resistive Pressure Sensor [40], (b) Capacitive Pressure Sensor [11], ( c) Optical Pressure Sensor [35]
Position and motion sensors are, widely, used in industrial projects. Positions can be detected by these sensors through different ways ranging from simple contact sensors to more complex and contact-free sensors. Location that is measured by the position sensors can be relative/absolute and linear/angular. Accelerometers, piezoelectric accelerometers, resistive and capacitive accelerometers are the popular position and motion sensors. Accelerometers are one of the motion sensors that measure the acceleration degree of objects.
The most accelerometers work based on the resistive and piezoelectric features. Piezoelectric accelerometers use the piezoelectric effects in a way that an electric charge is created whenever the sensing material is squeezed or strained by the sensor. These accelerometers are usually durable and, also, are protected from the contamination of noise influences. Resistive and capacitive accelerometers utilize an elastic cantilever with an attached mass. A force proportion to the acceleration deforms the cantilever whenever the sensor is considered to the acceleration objectives. Capacitive sensors compensate the cantilever unit in order to work as one electrode via an electrode strip. These sensors can be used to measure a constant acceleration (e.g., gravity on earth) [58]. Figure 3 illustrates various types of the position and motion sensors.
Figure 3. Three types of the Position and Motion Sensors, (a) Accelerometers Sensor [2], (b) Piezoelectric Accelerometers Sensor [39], (c) Resistive and Capacitive Accelerometers [43]
Temperature sensors measure temperature of various environments through a changeable mechanism into a physical device (e.g., resistance or output voltage). Electromechanical, electronic, and thermo-resistive are the most popular types of temperature sensors. Electromechanical temperature sensors work based on the expanding or contracting features of some materials on climate changes. The expansion rate generates an electromechanical motion when the material indicates a temperature change. Thermistor sensors are the resistors which contain different resistances. They are generally composed of a combination of two or three metal oxides that are sintered in a ceramic material. Thermistors are classified into two categories: Positive Temperature Coefficient (PTC) and Negative Temperature Coefficient (NTC). PTC devices display an increase in the resistance as temperature rises. In contrast, NTC devices display a decrease in the resistance via enhancing the temperature value. Resistive temperature detectors are developed by pure metals, such as copper, nickel, or platinum [62]. Figure 4 shows the three temperature sensors that are used in the industrial products.
Figure 4. Three instances of the Temperature Sensor [7, 25, 42]
Humidity sensor measures the quantitative amount of water vapor in the desired substance (e.g., a gas). Humidity is an essential feature on the environments that can be used in different applications, such as room air humidity in patient monitoring, exhibit perseveration in museums, soil humidity in agriculture, and process control in industrial usages. Humidity can be measured by humidity sensors as the absolute humidity (e.g., ratio of water vapor), relative (e.g., the saturated moisture level), or dew point (e.g., temperature and pressure of the gas turning into liquid). The most popular humidity sensors work based on the capacitive, resistive, and thermal techniques [53]. Figure 5 depicts two instances of the humidity sensor, including Sensirion SHT1x and Digital Humidity Sensor SHT7x.
Figure 5. Two types of the Humidity Sensor [46, 47]
Chemical sensor measures the presence or concentration of the chemical elements or compounds. It is, typically, composed of a chemically sensitive film or a membrane and, also, a transducer. The chemical process happening in or on a chemically sensitive film or in a membrane generates a signal at the transducer. Chemical sensors can be used in various applications, such as medical diagnostics, nutritional sciences, and automotive industry. The most chemical sensors are classified into five categories: inter-digital transducer sensors, conductivity sensors, optical chemical sensors, ion sensitive FET sensors, and piezoelectric chemical sensors [ 27, 57]. Figure 6 shows a general type of the chemical sensor.
Figure 6. An instance of the Chemical Sensor [12]
Biosensor measures the presence and concentrations of the bacteria, viruses, or molecules. Furthermore, they can detect different types of the molecular complexes, such as proteins, enzymes, antibodies, and DNA. This sensor is composed of three parts to work based on the signal received from the transducer: sensitive layer, transducer, and electronic circuitry. It is worth to noting that the sensitive layer embedded in the biosensor is a biosensitive biological component, such as enzymes, antibodies, cell membrane receptors, or tissue slices [60]. Figure 7 illustrates a typical biosensor.
Figure 7. An instance of the Biosensor [10]
Applications of sensor networks are applied by industrial companies in the recent decades. A sensor network consists of the low-power, low-energy, and low-cost sensor nodes. The sensors sense phenomena data (e.g., accelerator) and, then, transmit the sensed data to a center (e.g., sink).
They can be propagated in various environments under three scenarios: random, planned, and grid. Any strategy, offered to sensor networks, should reduce the device complexity and energy consumption to prolong the network lifetime. Environmental monitoring, healthcare systems, and military usages are the most popular applications of sensor networks [ 6, 14, 18-20, 22, 23, 31, 61]. Figure 8 illustrates a sensor network which detects phenomena data by several sensor motes.
Figure 8. A Sensor Network composed of Various Sensor Motes
Environmental monitoring is one of the popular WSNs applications that can be implemented by sensor networks. In this case, sensor nodes will be propagated in various areas to measure the meteorological and hydrological parameters (e.g., temperature and humidity). Applications of environmental monitoring are generally categorized into two groups: indoor monitoring and outdoor monitoring. Indoor monitoring applications, usually, surveillances buildings and offices based on the temperature, light, humidity, and air quality. Outdoor monitoring applications are typically used in habitat monitoring, traffic monitoring, earthquake detection, agriculture usages, chemical hazardous detection, volcano eruption, flooding detection, and weather forecasting. Sensors and actuators play a key role to perform the monitoring tasks.
Autonomy, reliability, robustness, and flexibility are some of the required requirements in monitoring systems. Autonomy indicates that energy batteries should able to supply the base stations along the whole deployment phase. Since the radio transceiver of sensor networks is a massive energy consumer, the network should be energywise under different environmental conditions.
Reliability represents that the network conducts the simple and predictable operations to prevent all or some of the unforeseen and complex problems. Therefore, human maintenance can be prevented since end-users may not have the networking knowledge and, thereby, various environments are, often, monitored remotely. Robustness indicates that the network should be stable in face of the most occurred problems (e.g., poor radio connectivity). Flexibility represents that the system operator should able to quickly add, move, or remove any desirable sensor node or stations at anywhere and anytime according to the demands of developed applications. Hence, sensor nodes should automatically identify their neighboring nodes to be flexible under such changes. All of the above requirements are essential, especially when a network is deployed in the remote and difficult-to-access places [34]. Figure 9 shows schematic of a fire system for forest monitoring.
Figure 9. Schematic of a Forest Monitoring System
Since healthcare topics contain the quality of life, researchers and scientist attempt to find appropriate procedures to facilitate their operations. Healthcare systems can be implemented by WSNs applications to enhance the performance of medical operations. Wireless sensors make different effects on the physical, physiological, psychological, cognitive, and behavioral processes in the various fields ranging from personal to large-scale buildings. Such effects can be achieved in sensory information based on healthcare requirements. There are some healthcare applications, such as monitoring in disaster areas, vital monitoring in hospitals, at-home and mobile usages, assistant via motor and sensory decline, and large-scale medical studies that can be designed and implemented by sensor networks. These applications indicate one of the medical requirements which are considered by heath operations.
Like any technology-based system, WSNs applications for healthcare systems involve some problems and also difficulties that should be considered by researchers carefully. Trustworthiness, privacy and security, and resource scarcity are the major challenges in these systems. Trustworthiness indicates the essential requirements for the end-to-end reliability. Furthermore, it considers the end-users' demands to measure the accurate data in medical operations. There are various factors (e.g., signal interference in the presence of metal doors) to complicate capabilities of the systems which offer the trustworthiness of medical systems. Therefore, a sensor network for medical purposes should use various techniques, such as automated data validation, cleansing, and interfaces. Privacy and security represent that WSNs applications cause to different problems for privacy violation. This process is mentioned to define possibility of the daily living and, also, make adequate information for longitudinal studies. Attenuating the signal outside of the home, transmitting radio messages in the period of time, delaying the radio messages randomly, hiding the fingerprint of transmitter, and transmitting the fake data are some of the suggested solutions for the above problems. Resource scarcity of sensors uses lowpower components with the constrained resources. The limited computation, communication, and energy resources of sensor nodes cause various problems to be occurred in healthcare systems. Physiological monitoring, motion and activity monitoring, and large-scale behavioral studies are some of the wireless sensing prototypes that are developed for medical operations [29]. Figure 10 illustrates schematic of a healthcare system. Each human and object carries a sensor mote to transmit the sensed data (e.g., blood pressure) to the center placed at the hospital.
Figure 10. Paradigm of a Healthcare System
Communication technologies are applied by military usages to transmit messages between incident commander and team members. Military operations can be developed by WSNs applications to achieve some of the communication purposes and ensure the distribution of logistical and intelligence data from sensors. Some of the sensors, such as chemical, biological, radiological, nuclear, and explosive are used in military usages under four scenarios: battlefield, urban warfare, other-than-war, and force protection.
Sensor networks develop military applications by using six classes: self-healing land mines, soldier detection and tracking, early attack reaction sensor, sniper detection and localization, perimeter protection, and sound recognition system.
Self-healing land mines indicates that an antitank mine surveillances the situations of its neighboring nodes, detects the main threats, and responds to the threats by a moving action autonomously. The sensing process is controlled based on the accelerometer sensors and also distributed, self-contained, acoustic location system. In the soldier detection and tracking class, desirable places (e.g., military sites and buildings) are monitored to detect the enemy individual soldiers. To obtain these goals, the composition of acoustic sensors and daylight still cameras is used by a WSNs application to monitor the human tracking. Early attack reaction sensor is a passive, acoustic sensing system to measure the gunshots on monitored areas. It should transmit the relative azimuth and range information of the shot location toward an assigned operator. In the sniper detection and localization class, the acoustic localization of shots is used to increase the soldier's protection against snipers. It uses the mobile antennas which are installed on the soldier's helmet. Perimeter protection uses a multi-sensor system (e.g., day/night cameras and millimeter-wave radars) to measure the radiation waves which are reflected from the defined targets. In the sound recognition system, a sound detection mechanism is effectively applied by the asymmetric warfare and against terrorist threats.
The military applications that are designed and implemented by sensor networks should consider two network parameters: data reliability and denial of service. Data reliability represents that data should be successfully transmitted to the end-users. Furthermore, data should be delivered to the incident commander or team forces through a secure way without any opportunity to the interception and tampering by any eaves-dropper. Denial of service indicates that any WSNs application for military operations should able to react against the denial of service created by an adversary. This goal can be achieved by reporting the incident of an attack (e.g., jamming) [17]. Figure 11 depicts schematic of a military war system. Each team member and vehicle system equips a sensor mote to transmit their status toward the incident commander.
Figure 11. Schematic of a Military System
Location of sensor nodes is required to both network operations and WSNs applications. Therefore, sensor localization is one of the main operations in sensor networks. Since the financial costs of Global Positioning System (GPS) devices are high as well as GPS signals are not available in the enclosed environments, GPS-based localization cannot be used in large-scale sensor networks. Consequently, researchers have presented other localization methods (e.g., spatial relations of nodes) that do not require any specialized hardware and/or do not employ only a limited number of awareness anchors.
As shown in the illustration of Figure 12, localization methods can be categorized into three groups: proximity-based localization, range-based localization, and angle-based localization. Proximity-based localization is used in terms of a graph model. The proximity measurements are specified by the two famous models: adjacency matrix and distance matrix. Rangebased localization indicates that the range of wireless signal transmissions can be measured to determine the sensor location.
Figure 12. Three groups of the Sensor Localization in WSNs Applications
Since sensor nodes are equipped with radio hardware for serving the defined communications, the distance among nodes can be estimated by the received signal strength. Angle-based localization applies the extra angle measurements to determine the sensor location. Since the network needs to the antenna array or multiple receivers on nodes for the angle-based localization, the financial costs of this localization method are very high [ 45, 48, 55].
Data mining techniques discuss about inference process of the knowledge from the large data which are gathered by WSNs applications. Because sensor data contains specific characteristics and sensor networks have some natural constraints, traditional data mining techniques cannot be directly used in WSNs applications. Data mining techniques are categorized into four groups, as shown in Figure 13: frequent pattern mining, sequential pattern mining, clustering, and classification. Frequent pattern mining is used to look for the group of variables which co-occurs frequently in the dataset. The major objective of this technique is to find the most interesting relations among variables. Frequent pattern mining is developed to find more complex structures (e.g., sequential pattern mining). It displays frequent instances as patterns in the sequence database which stores a number of records. This process is done where all of the records are sequences of ordered occurrences. Sequential pattern mining presented by researchers in the recent years can be used in various applications such as medicine, local weather forecast, and bioinformatics. Clustering technique is derived from an unsupervised learning in a way that a given data is indicated by subsets as well as each subset indicates a cluster having the distinctive properties. It is an appropriate technique especially for the WSNs applications which need scalability for the large number of sensor nodes. Furthermore, it supports data aggregation to summarize the whole transmitted data. Classification technique is considered as a task to assign new objects into a class of the pre-defined object categories. It is learned by using the set of training data as well as it can classify new data into one of the learned classes. All of the above techniques use the centralized methods to solve application-based issues, the centralized procedures to maximize the network performance, and the distributed mechanisms to improve the application-based issues and network performance [32].
Figure 13. Four Groups of the Data Mining Techniques in WSNs Applications
This section evaluates the performance of Wireless Sensor Network (WSN) compared to RFID network in terms of utilization, purpose, components, mobility, power supply, programmability, and deployment. Furthermore, some of the WSNs simulators are compared to each other in terms of computation time and memory usage. RFID system enables the identification operation into a defined range and operates without need to any sight line. RFID tags accept a larger set of unique IDs than bar codes. RFID systems can utilize many different tags that are located in the same global area without any human assistance. There are many types of RFID systems, but at the highest level, RFID devices are categorized into two classes: active and passive. Since active tags need a power supply source, they are connected to either a powered infrastructure system or the power energy stored into an integrated battery. A passive tag is composed of three parts: an antenna, a semiconductor chip on the antenna, and some form of encapsulation [16, 28].
Figure 14 shows the performance analysis of the RFID and WSN systems based on the utilization rate [1]. The comparison process is carried out via the Automated Storage and Retrieval System (ASRS) and the three machines, including Machine 1, Machine 2, and Machine 3. MATLAB Petri net toolbox [33] is used for the performance measurement. There are 29 places and 22 transitions in the case of RFID scenario as well as there are 28 places and 18 transitions in the case of WSN scenario. Furthermore, ultra-high frequency (UHF) RFID tags are used in the simulation process. The comparison results indicate that the utilization of RFID is higher than that of WSN in the ASRS and all the simulated machines. Table 1 represents various difference between WSN and RFID in terms of purpose, components, mobility, power supply, programmability, and deployment [54]. WSN senses the parameters which define environmental situations of the objects or environments (e.g., temperature). In contrast, RFID detects and identifies the tag objects which are attached on humans [36, 37], vehicles, etc. WSN uses sensor nodes, relay nodes, and sink; while RFID uses the transponder and interrogators. Sensor nodes are usually static, but RFID tags move with the attached objects. WSN is equipped to a battery to supply the energy of sensor nodes; while RFID tags are battery powered or passive.
Table 1. Differences between WSN and RFID [54]
Figure 14. Performance Evaluation of the RFID and WSN Networks based on Utilization Rate [1]
WSN is programmable based on the implemented application. In contrast, RFID is composed of nonprogrammable tags. Finally, sensor nodes are propagated in a random or fixed model, but RFID tags are deployed in a fixed model.
WSNs can be simulated by various network simulators such as OMNeT++ [38], NS-2 [59], and NS-3 [26] before the implementation stage. Figure 15 shows performance evaluation of the above simulators under two simulation scenarios [56]. Figure 15(a) illustrates the effects of network size on computation time. The comparison results indicate that computation time of the NS-3 simulator is low compared to the other simulators. Therefore, the efficiency of this simulator is better than OMNeT++ and NS-2. Figure 15(b) depicts the effects of drop probability on memory usage. The evaluation results display that the memory storage consumed by the NS-2 simulator is more than those consumed by the OMNeT++ and NS-3 simulators. That is, NS-2 is not efficient, in this simulation scenario, compared to the other simulators.
Figure 15. Performance evaluation of the WSNs simulators [2]. (a) Computation Time vs. Network Size, (b) Memory Usage vs. Drop Probability
Wireless Sensor Networks (WSNs) are deployed by the lowpower, low-cost, and large-scale wireless sensor nodes. The nodes measure environmental conditions, including temperature, humidity, etc. Afterward, they forward the sensed data to a center (e.g., base station). This paper, firstly, has described various physical features of the pressure, position, motion, temperature, humidity, biosensor, and chemical sensors. Secondly, it has presented some of the most popular WSNs applications, including environmental monitoring, healthcare systems, and military usages. Besides, the paper has discussed on different techniques of the sensor localization and data mining in sensor networks.
In the performance evaluation, Wireless Sensor Network (WSN) has compared to radio-frequency identification (RFID) in terms of utilization, purpose, components, mobility, power supply, programmability, and deployment. The comparison results indicated that the utilization rate obtained by RFID network could be increased by nearly 20% more than that obtained by WSN network. Moreover, some of the WSN simulators, (including OMNeT++, NS-2, and NS-3) have compared to each other in terms of computation time and memory usage. In the comparison results, performance rate of the NS-3 simulator was better than the OMNeT++ and NS-2 simulators, in the most cases.
Since the most users cannot utilize all of the sensor types in WSNs applications properly, it is recommended that they apply sensor modules in the implemented circuits. Moreover, it is suggested that industrial applications to be implemented by the integration of sensor networks and RFID networks because both of these networks have powerful advantages. Simulation results indicate that performance of NS-3 simulator is better than the OMNeT++ and NS-2 simulators in terms of computation time and memory usage. Consequently, the author recommends that users conduct simulation process of the WSNs applications (e.g., routing and congestion control) by NS-3 simulator.