The wireless sensor networks

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.


Advances in wireless communications during the last decade made the development of services based on wireless sensor networks easier and cheaper. Such a system is implemented where environmental and liquid flow data are sensed and relayed to the central computer via an ad hoc network based on the IEEE-802.15.4 standard. This paper presents the architecture of the hardware and the software platform designed to measure key properties of a sugar bush vacuum pumping system, relay data and detect then locate anomalies. The architecture is composed of three distinct sections: the sensor nodes, the wireless ad hoc network and the central application which process all data received from the nodes. The nodes being non hierarchical, permit a more rapid deployment, an easier maintenance and a better compatibility with various vacuum systems. More over the system is designed to be power efficient since each node spend most of the time in the sleep mode.


The tremendous progress made in recent years in micro electronics have made possible the realization of more powerful and less costly electronics. The wireless sensors are no exception, containing a sensing device, a computing unit and a wireless module, their performance are always improving while the price is constantly decreasing. The Wireless sensor networks (WSN) are commonly deployed in large areas [1], [2], the number of sensor nodes in the WSN is consequently very high and each sensor node is battery powered. Therefore the individual node cost must be very low and must require as low maintenance tasks as possible during the network lifetime [3]. The cheaper the wireless node the more limited capabilities it has in term of memory, processing capabilities and coverage range [3]. These constraints in addition to the fact that every node is battery powered make the design and the implementation of a WSN a delicate and challenging task. The WSN cover a very large area of applications such as the industrial and security applications, medical applications, commercial ones, in addition to the environmental and agriculture applications [1], [4], [5]. This paper is about the design and the implementation of a WSN for an agricultural application.

State of the art

Sensors have been employed for agricultural application over the past twenty years [6]. The early adopters for the sensors technology found it unprofitable since the electronic components used to be more expensive and less reliable than it is today. Therefore this kind of technology was restraint to largest farms [4]. The sensors deployed at this time used to be either wired or unplugged to any communication device so the user used to collect the data on the field by him self periodically. Moreover the data collected on the field at that time used to be moderately precise. The technological advances of the micro electronics led to lower prices with better performances, the use of WSN in agricultural applications have become easier despite the harsh environment and the very large covered area. WSN agricultural applications deployed many kinds of network architectures (i.e. infrastructure dependent networks [7], hierarchical networks [7], ad hoc networks [4], [8] ...).

In addition to these different kinds of architectures, several wireless technologies are available with different Physical (PHY), Medium Access Control (MAC), and Network (NWK) layers specifications. The most commonly used standards for this kind of technology are the Bluetooth standard [9] and the ZigBee standard [10]. The Bluetooth standard is a specification for wireless personal area network (WPAN), its range is typically 10 meters but 100 meters in line-of-sight (LoS) communication could be ensured with a 100mW transmit power. It operates on the 2.4 GHz industrial scientific and medical band (ISM). The Bluetooth signal occupies an instant band width of 1 MHz. Being based on the frequency hopping spread spectrum technique (FHSS) with 1600 hops per second and 79 centers frequencies, the total occupied bandwidth of this technology is 79 MHz for a bit rate of 1 Mbit/s for the standard's version 1.2 [9]. Therefore the Bluetooth standard is more suitable for small ad hoc networks configuration needing a fairly high bit rate. On the contrary the ZigBee standard is a lower power consumption standard with a lower bit rate; it is more suitable for larger networks. This standard operates in both of 2.4 GHz and 868/928 MHz ISM bands it's maximum bit rate vary from 250 kbps to 40 kbps and it's typical range is 30m for non line-of-sight (NLoS) use and 100m for LoS, but newer enhanced ZigBee modules can reach 140m NLoS and 3000m LoS with only 50mW transmit power. The ZigBee Protocol defines three types of nodes: Coordinators, Routers and End Device, with a requirement of one Coordinator per network [10]. The figure 1 shows a typical ZigBee network architecture.

While all nodes can send and receive data, there are differences in their specific roles. Coordinators are the most capable of the three node types. There is exactly one coordinator in each network; it is the device that establishes the network originally. The coordinator is able to store key information about the network, such as security keys. Routers act as intermediate nodes, relaying data from other devices. The coordinator and the router are commonly called the Full Function Devices (FFD) while the end devices are commonly called Reduces Function Devices (RFD). RFD are low power consuming devices and are usually battery powered. They have sufficient functionality to talk to an FFD but cannot relay any data. These Devices are the only nodes capable to enter in the sleep mode [10]. Therefore if the FFD are battery powered with the same battery type of the RFD, the FFD, consuming more power, will run out of energy long before RFD. If one of the FFD devices fails, a whole cluster of the network will fail, therefore a more homogeneous network would be more suitable for large area covering and battery powered sensor networks.

Problem statement and proposed solution

Problem statement:

Canada has been and is still the first maple syrup producer in the world [11] with 4.9 million gallon in 2008 [12]. The production process has been constantly improved with the technological advances, from the traditional production using buckets to an efficient and automated technique by using a vacuum pump [11], reducing labor and processing costs [11], [13], [14]. The tubing of the vacuum system is vulnerable to falling trees, bottlenecks and rangers. If such a problem occurs on any conduit in the sugar bush, the system performance will drop especially if the damaged tube is a large section one. The exploiter will have to search in a very large, hard to reach area and in a harsh climatic conditions in order to find the problem and fix it. This will lead to a loss of time as well as human resources and maple syrup. A monitoring system capable of detecting and locating any problem occurred in the tubing then reporting it in real time to the exploiter would reduce labor and processing costs in addition to enhancing the production. Since a wired system will have the same weak points (falling trees and rangers) as the tubing, a WSN capable of detecting and reporting any dysfunction to a central unit accessible to the exploiter would be a great solution. In order to maximize the profitability, robustness, and flexibility of the system, the following constraints must be satisfied.

  • Cost-effectiveness.
  • Low power consumption for longer lifetime.
  • Expendability to other applications with no or minor adjustments.
  • Long range and non line of sight Communication.
  • Optimized system against interference and noise.
  • Easy deployment and maintenance.
  • Friendly and intuitive user interface.

The satisfaction of these constraints require an effort in both software and hardware level.

System description:

In this paper, a WSN based application has been designed to monitor the flow of the maple syrup in a sugar bush tubing as well as the ambient temperature. This is made possible by a pressure sensor and a temperature one. With both of the temperature and the pressure data the computing unit can decide weather the system performs in the optimal conditions and if a problem occurred. The temperature and pressure data will be sent from the sensor node to the user either periodically or upon user request. Figures 2 and 3 illustrate such scenarios.

The data collected from both of the sensors is digitalized then packetized, ready to be sent via a wireless module then routed on the ad hoc network to reach the control unit (user). This data is used by the control unit to show the sensors status. If a problem had been detected in one or more nodes, the system will check the remaining nodes in the area and draw a map with the estimated location of the problem. When the location is known the user could go directly to the indicated area to fix the problem.

Sensor node design:

As shown in figures 2 and 3, the designed system is a set of sensor nodes connected the each other via a radio link. The whole WSN is connected to the control unit via a sink node. This particular node performs as an interface between the control unit and the WSN. All the nodes in the WSN except the sink node are battery powered and have exactly the same architecture; this architecture is described in the figure 4.

The measures from sensors are periodically performed by digitalizing their analog response. The measures period is adjustable and could even reach 15000 samples per second. In our application there is no need to use such a high sampling frequency, but this could be useful for other applications since our system is designed to be easily customized for other needs. The analog-to-digital converter (ADC) used is a 10 bits accuracy converter embedded in the used microcontroller (µc) and running in the noise reduction mode. When a conversion request is received the processing unit stops the central processing unit (CPU) and all of the input/output (I/O) modules such as the Universal asynchronous Receiver Transmitter (UART). The (I/O) modules and the integrated asynchronous timer needed for the conversion are not stopped. This improves the noise environment for the ADC, allowing higher resolution measurements with more accuracy. The 10 bits accuracy with the sensors output voltage range ensures a good precision for both of the pressure and the temperature sensor. The structure and the characteristics of both of the sensors are respectively described in figures below.

As shown in the figure 5 the used pressure sensor is not only a sensing element but it's also equipped with two level gain stages to ensure a high level output signal with temperature compensation providing more accurate measures. The transfer function of the used pressure sensor is shown on figure 6.

The sensor is able to measure vacuum pressure values from -7.2519 PSI to 0 PSI which is the atmospheric pressure. It delivers an analog voltage proportional to the measured pressure voltage varying from 0.2V to 2.8V as shown in the figure 6. The expression of the output voltage, according to the constructor, is:

Vout = VS (0.018 P + 0.94) ± (0.0225 VS). Where P is the actual pressure and VS is the sensor's supply voltage.

As shown in the figure 7 the used temperature sensor provide not only a proportional analog voltage Vout to the temperature measured on SD but also the possibility to get a hardware critical temperature interruption with the thermostat comparator outputs TO and . like the pressure sensor, the temperature sensor delivers an analog voltage representing the measured temperature. The output voltage characteristic of the sensor is shown in the figure 8.

The temperature sensor is able to measure the temperature varying between -40°C and 120°C with an output voltage varying from 0.2V to 1.3V. This output voltage is converted, like for the pressure output voltage, to a digital 10bit word in order to be sent.

To ensure lower power consumption the sensors are powered on only when a measure is needed. Both of the sensors spend most of the time offline this contributes to the low power consumption strategy adopted on the µc and the wireless communication module by turning them periodically on the sleep mode. By spending most of the time on lower power consumption mode the battery lifetime of each node is expended which lead to the lifetime increase of the whole system. The diagram on the figure 9 describes the running program on the µc and shows how the system switches to the sleep mode after performing any task.

On startup, the µc configures himself by setting up the input and the output ports, the UART baud rate and the ADC conversion parameters. After the self configuration step the µc will check the wireless module's configuration and enter immediately on the sleep mode.

The sleep mode can only be interrupted by critical temperature or pressure detection or when the sleep period finishes. When the data to be transmitted is ready the µc will send a "request to send" to the wireless module. When ready the wireless module will respond with "clear to send" and the transmission will begin. This feature is particularly useful to prevent data loss which could be due to a transmit buffer overflow in the wireless module.

The wireless module used implements the 802.15.4 specifications in the 900MHz band, this enhances the module's immunity to the radio canal noise and increases the range achieved in the 2.4 GHz band by 4 to 6 times. The module implements also the routing protocol DigiMesh [15]. This Routing Protocol is based on the Well known reactive routing algorithm AODV, leading to a reduction of the routing overhead. This protocol is particularly suitable for our application since it doesn't need any infrastructure deployment, and any node can enter and leave the network without causing the network as a whole to fail. The network allows all the nodes be on the sleep mode with a synchronized wakeup [15], the figure 10 shows the architecture of the DigiMesh network.

While the ZigBee protocol defines three types of nodes [10], the DigiMesh protocol defines only one [15] therefore the network is perfectly homogeneous. The network does not rely on a particular node and the failure of one node does not affect the whole network. This characteristic offer more simplicity on deployment and more flexibility for possible expansion. All the nodes have the same embedded running program, therefore they will do similar tasks and will have almost similar lifetime. The figure 11 describes the running program on each of the nodes.

When powered on, the wireless module (XBee module) enters to the Sleep mode. When a "request to send" is received, the module will respond with a "clear to send" message to the µc. The µc will then begin to send the data to the XBee module which will buffer the received data and try to discover the destination node. If the destination node is found the XBee module will send the data to the next module in the discovered route. The modules will relay the packets until the destination is reached.

System's software design:

The application's user software is designed so that the following constraints are satisfied:

  • Friendly and intuitive user interface.
  • Modular and generic design for more flexibility and easier extendibility.
  • Display every node status with the newest available data.
  • Store all the received data for later history display.
  • Estimate the problem location and display it in a map.
  • Generic application which could be used for other applications requiring measurements and data processing.

The application is composed of independent units. Every unit provides, with its own resources, a full service to another one. The advantages of such architecture are the ease in the design of the application and easier modifications. Moreover changes on any unit is possible without modifying any other unit and without affecting the system's operation

To ensure more stability to the system, the application runs several threads, each thread operates independently. This allows the application for instance to receive the data from the serial connection while updating in real time the displayed measurements, saving the history and listening for new events.

The figure 13 illustrates a simplified flow diagram of the main tasks executed in the user application. The application contains other functionalities not represented in the flow diagram, such as the displaying of the measures and the statistical data. These functionalities are not represented in order to simplify the diagram in the figure 13.

System performances

In this section, several tests and measures are performed to evaluate the system's performances when deployed in the real application. The accuracy of the measures performed by both the temperature and the pressure sensors are first evaluated. Then the impact of the baud rate on the delivery rate is studied. For a particular baud rate the range and the delivery rate of the system are measured and commented. Finally, to evaluate the power consumption performances, a simulation is executed using a wireless sensor network simulator developed specially to meet the requirements of this application [x].

Accuracy of the measures:

The developed application has been used to generate the curves which describe the variation of the measured temperature and pressure in the ambient temperature and under the atmospheric conditions. The figure 14 and 15 below describes respectively, the variation of the sensed temperature measured in Celsius and the evolution of the pressure expressed in psi.

In order to measure the accuracy of the system, the laboratory has been closed and set to constant temperature and the sensor has been placed in a thermal insulation box. The wireless transceiver as well as the microcontroller and the power management system, has been kept out in the ambient air in order to avoid the heating effect caused by these equipments in the thermal insulation box. Since the laboratory has been closed and kept in a constant temperature, the pressure variation is almost null. Being thermally insulated and under constant pressure, the accuracy of the temperature, as well as the pressure sensing capabilities, are negligibly affected by the ambient air temperature and pressure changes. Thus the variation of the reported temperature and pressure, shown respectively in the figure 14 and 15 in green, are due to two essential factors; the sensor's intrinsic measurement errors and the analog to digital converter's noise. The measured temperature range is (zzz Max-Min) with a maximum value of zzz and a minimum value of zzz, while the measured pressure is between (zzz) and (zzz) i.e. within a (zzz) range. The application delivers a mean temperature almost equal to (zzz) while the mean pressure is around (zzz). The standard deviations expressed in the figure 16 and 17.

(zzz figures standard deviation)

As shown in the figure 16, the standard deviation expressed in °C is ..(zzz) while the standard deviation of the pressures measurements depicted in the figure 17 is ..(zzz). Dividing the standard deviation by the mean of the measures gives an idea about the system's precision. This is shown in the figure 18. The (zzz) curve shows the precision of the temperature sensing, while the (zzz) one expresses the accuracy of the pressure sensing.

Impact of the Baud rate on the delivery rate:

Range test and delivery rate:

The range tests have been made with a wireless module (emitter) sending 32 Byte of data (power) and receiving them back from the second node (receiver) (power) with a loopback connection, making it acting as a mirror. The experimental setup is depicted in the figure 15.

The range of the developed sensor nodes is a crucial parameter since the WSN is to be deployed in a very vast area in a very harsh environment. Three range tests have been performed, the first have been done during a snow storm to show the system performance in such weather conditions. The second test have been made in similar conditions of the first test except the fact that the receiver wireless sensor was covered by about 0.5cm of snow layer. The third one have been done in a wood containing conifer as well as Deciduous trees with 2 to 4cm of snow layer covering tree branches. The height of the emitter as well as the height of the receiver is 1.5 meter. The height of the trees is between 30 and 40 meter. Thus the experimental conditions are very similar to the real application's conditions. These heights make the signal totally submerge in the forest environment, which is noisome to the electro-magnetic waves.

The typical received power of both wireless modules varies between -78dBm and -112dBm. In the results discussions, the 100% value expresses a -78dBm received power or above and the 0% value refers to a -112dBm received power or below.

The red line in the figure 16 shows the path followed at the range test. The yellow mark in the upper corner of the satellite image marks the beginning point while the mark in the lower left corner marks the end of the measures. The figure 17 shows the first test results.

The figure 17 illustrates the system's delivery rate and signal strength evolution with the distance varying from 0 to 420m and the elevation varying from 14 to 9m over the sea level. The system appears to be perfectly reliable until 270m. With nearly 100% delivery rate and 80% signal strength. Despite the low performances between 295 and 310, which could be due to the fact that the emitter was masked by the descent, the system performance is slightly better for the next fifty meters. Then it drops again for the last sixty meters.

The figure 18 below represents the path followed in the second experience with the wireless module enclosure was covered with a thin (~0.5cm) layer of snow. The yellow pin in the right represents the starting point of the experience; the other one represents the end. The results of this experience are depicted in the figure 18.

A third experience have also been done in the wood, the followed path, the upper yellow mark marks the end of the experience while the other one marks its beginning. The path followed in this experience is shown on the figure 20 while the results are shown in the figure 21.

In the third experience, the system shows excellent performances despite being in a very harsh environment. In fact the wood and the snow are very harmful to the electromagnetic wave. The system behaves perfectly well until 110m with fair performances until 150m which is largely enough for the targeted application.

Power consumption and network lifetime estimation:


Conclusion and Future work

In Conclusion, the system's (performances ... zzz) range is perfectly acceptable for the targeted application. The whole concept could be used for different monitoring applications with minor adjustments. The developed solution meets the constraints fixed above; It is generic and flexible since the use of different sensors is straightforward and the user application is portable to any platform. Nevertheless other tests must be done to conclude about the sensor node effective power consumption. Moreover the application has to be tested also in a real sugar bush to conclude about its performance.

Several enhancements could be envisaged to improve the system's performance. In fact a non invasive sensing method is being studied; this method is based on the ultrasonic time of arrival technique. The non invasive sensor will certainly enhance the solution's flexibility since the nodes could easily be moved to other locations without any additional effort.