Ittipong Khemapech

*University of the Thai Chamber of Commerce, Thailand* 

#### **1. Introduction**

16 Will-be-set-by-IN-TECH

476 Environmental Monitoring

C. Alippi, R. Camplani, G. Galperti, and M. Roveri, "A robust, adaptive, solar-powered WSN

J. Bazerque and G. Giannakis, "Distributed spectrum sensing for cognitive radio networks by

S. Kim, K. Koh, M. Lustig, S. Boyd, and D. Gorinevsky, "An interior-point method for

M. Figueiredo, R. Nowak, and S. Wright, "Gradient projection for sparse reconstruction:

H. Zhu, A. Cano, and G. Giannakis, "Consensus-based distributed MIMO decoding using

Selected Topics in Signal Processing, vol. 1, pp. 586–597, 2007

semidefinite relaxation," In: Proceedings of CAMSAP, 2007

45–55, 2011

Processing, vol. 1, pp. 606–617, 2007

2010

framework for aquatic environmental monitoring," IEEE Sensors Journal, vol. 11, pp.

exploiting sparsity," IEEE Transactions on Signal Processing, vol. 58, pp. 1847–1862,

large-scale -1 regularized least squares," IEEE Journal of Selected Topics in Signal

application to compressed sensing and other inverse problems," IEEE Journal of

Energy conservation is currently growing in importance. This chapter focuses on the issue of energy conservation within the domain of Wireless Sensor Network (WSN). There are also specific reasons why energy conservation is more important for WSN than for other types of networks. A WSN consists of multiple sensors which are able to sense some aspect of their environment and communicate their readings to a base station or sink without being physically connected to it. Sensors are often also resource constrained, being small in size and relying on small batteries for power. Consequently, the efficient utilisation of energy should be an important priority for designing WSN network protocols. This differs from the traditional approach to designing network protocols where issues like survivability, maximising throughput or reliability have been prioritised. Making energy conservation an important design priority is a new approach.

Wireless sensor network (WSN) is an important research area with a major technological impact. With significant breakthroughs in "Micro Electromechanical Systems", or MEMS, (Warneke & Pister, 2002), sensors are becoming smaller. It is feasible to fit them into a smaller volume with more power and less production costs. WSN may be deployed in a wide range of different environments. These include remote and hostile environments as well as local and friendly ones. A major driving force behind research in WSN has been military and surveillance applications. Recently, however diversification has occurred with the development of civil applications. One example which is used as a reference point throughout this work is Great Duck Island (GDI). Sensors were scattered over a remote island to monitor the seabird's migration (Mainwaring et al., 2002). In another example WSN was deployed around volcanoes (Allen et al., 2006). Such applications illustrate the usefulness of WSN which make data collection feasible from remote and hostile environments with minimal human intervention.

One of the main objectives of WSN power conservation is to minimise energy usage whilst other functional requirements such as reliability or time synchronisation are still achieved. Some authors argue that multi hop communication allows for deployment in scenarios where direct communication with a base station is not practical (Arora et al., 2004; Allen et al., 2006; Chintalapudi et al., 2006). However, the spread of the Internet means that wireless devices may often communicate directly with a device that is connected to the Internet and has a reliable power supply. This work focuses on the design of wireless sensor networks protocols where direct communication with a powered base station is feasible and data is sent from the sensors to the base station at regular intervals. There are several important scenarios where such two assumptions hold.

former and latter cases, respectively. To be application specific results in a more complicated

In total, seven groups of applications have been categorised by us based upon their functionalities including habitat monitoring (HM) (Juang et al., 2002; Mainwaring et al., 2002; Szewczyk et al., 2004), environmental monitoring (EM) (Allen et al., 2006; Martinez et al., 2005), health monitoring (HEM) (Jovanov et al., 2003, Otto et al., 2006), structural health monitoring (SHM) (Chintalapudi et al., 2006; Kottapalli et al., 2003; Paek et al., 2005, Schmid et al., 2005), event detection and tracking (EDT) (Arora et al., 2004; Dreicer et al., 2002; Simon et al., 2004), transport monitoring (TM) (Coleri et al., 2004) and location-aware system (LAS) (Brignone et al., 2005). Specific capabilities and underlying communication paradigms have been outlined. For example, data encryption may be required in some health monitoring systems for transmitting a patient's diagnosis data to the main server located at the hospital. Furthermore, data correctness is also required in this case. In some applications such as event tracking and detection systems, several intermediate nodes are required for forwarding the sensed data to the base station. However, a direct communication from source to base station is found in some health monitoring systems. This section addresses application specific characteristics of WSN applications by detailing the differences in their

The "Event/Periodic Based" aspect demonstrates how often data reporting is conducted. There are three main types including event-based, periodic-based and hybrid. Each sensor is triggered to operate by the occurrence of an event in the case of an event-based application. An example of this application type is the Event Detection and Tracking. Congestion is one of the major concerns designing a protocol to support event-based networking as it is caused by a lot of traffic generated by all sources in an event area. The key idea of congestion avoidance is to control data reporting rate of such sensors (Sankarasubramaniam et al., 2003). However, the main assumption is that all data packets have the same priority. Packet loss is therefore tolerantly acceptable. There are several works on congestion control specifically developed for WSN (Ee & Bajcsy, 2004; Hull et al., 2004, Lu et al., 2002, Wan et al., 2003). The congestion control approach focuses on channel monitoring to dynamically adjust the data forwarding rate. CODA (Wan et al., 2003) has been designed to cover two types of problems corresponding to the deployed sensors and their data rate. However, it does not provide any queue occupancy monitoring. Sending an ACK (Acknowledgement) in the case of persistent congestion, even if it is small in size, may increase the number of traffic. This mechanism also requires feedback signalling which results in higher cost. Only packet prioritisation could be found in (Lu et al., 2002). However, it proposes the VMS (Velocity Monotonic Scheduling) policy which supports both static and dynamic computation of the requested velocity and it also solves the fairness problem. Both channel and queue occupancy monitoring are provided in (Hull et al., 2004) and (Ee & Bajcsy, 2004). A child node can transmit packets only when its parent does not experience congestion problems and some help from the MAC (Medium Access Control) layer to shift the transmitting time to avoid interference are proposed in (Hull et al., 2004). A similar concept also exists in (Ee & Bajcsy, 2004) by

Each sensor periodically performs its operation. Some examples of data collected by the sensors are temperature and humidity. The significant change in readings may be used to

design process, especially in the case of designing a generic power-aware protocol.

requirements.

**2.1 Event/periodic based** 

comparing the normalised rate of a node and its parents.

This research work specifically looks at WSN where direct communication is possible and beneficial. A protocol for WSN, Power & Reliability Aware Protocol (PoRAP), is developed and provides energy efficient data delivery, without increasing packet loss. In designing PoRAP several experiments were conducted to establish the relationship between transmission power, reception signal strength and packet reception success. These showed a strong correlation between Received Signal Strength Indicator (RSSI) and Packet Reception Rate (PRR). In PoRAP, the RSSI is monitored at the base station. If the RSSI is too high the base station signals the sensor to reduce its transmission level, thereby saving power. If the RSSI is too low the base station signals the sensor to increase its transmission level so that packet loss is avoided.

PoRAP adopts a schedule based scheme for the sources' transmissions. It is assumed that nodes will be reporting measurement data regularly back to the base station. Each reporting interval consists of three time periods. In the first the base station sends a configuration packet. This informs nodes whether they are to increase, decrease or leave unaltered their transmission levels. There are then slots, each of which contains a data slot within which a sensor may transmit its data to the base station. There may then be a period of quiet before the start of a new cycle. Delays and clock drifts are measured so that nodes know when to wake up to listen and transmit. Delays depend upon payload size.

The design aims to optimise energy conservation rather than system throughput, in many sensing scenarios high throughput is not required. Sensors collect and report some physical data such as temperature and humidity. In such cases, the data reporting rate may be in minutes or hours. Two packet structures are used in PoRAP. The control packet is used in control and setup phase. It contains essential information for transmission power adaptation and scheduling. The data packet is used to deliver the collected physical data back to the base station.

The remaining parts of this chapter is organised as follows: Section 2 addressed application specific WSN. At present, WSN has been used in both military and civil applications. Each application category has particular characteristics and its own set of requirements. Hence, there are significant challenges in a generic protocol design for a variety of applications. Resource constraint issues are provided in Section 3. Apart from limited power resources, sensors also have constrained communication ranges for indoor and outdoor environments. The distance between the source and destination is crucial to employing an appropriate underlying communication paradigm. Section 4 describes the experimental details and their results which motivate the design of PoRAP. There are several factors which affect the link quality metrics such as distance between source and base station and time of day. The design of PoRAP is outlined in Section 5. PoRAP consists of several TinyOS components at the source and base station. The results shown in Section 4 motivates the design. The results of PoRAP evaluation in terms of energy conservation are presented in Section 6. Lower transmission power can be used to save the power whilst the reliability is in the desired range. Finally, Section 7 concludes the chapter.

#### **2. Application specific WSN**

Apart from being used in military or surveillance, WSN has been deployed in several civil applications which have different requirements. Periodic sensing is required in some habitat and environmental monitoring systems whilst event sensing is the norm in surveillance systems. Network lifetime and data reporting rates are therefore major concerns for the

This research work specifically looks at WSN where direct communication is possible and beneficial. A protocol for WSN, Power & Reliability Aware Protocol (PoRAP), is developed and provides energy efficient data delivery, without increasing packet loss. In designing PoRAP several experiments were conducted to establish the relationship between transmission power, reception signal strength and packet reception success. These showed a strong correlation between Received Signal Strength Indicator (RSSI) and Packet Reception Rate (PRR). In PoRAP, the RSSI is monitored at the base station. If the RSSI is too high the base station signals the sensor to reduce its transmission level, thereby saving power. If the RSSI is too low the base station signals the sensor to increase its transmission level so that

PoRAP adopts a schedule based scheme for the sources' transmissions. It is assumed that nodes will be reporting measurement data regularly back to the base station. Each reporting interval consists of three time periods. In the first the base station sends a configuration packet. This informs nodes whether they are to increase, decrease or leave unaltered their transmission levels. There are then slots, each of which contains a data slot within which a sensor may transmit its data to the base station. There may then be a period of quiet before the start of a new cycle. Delays and clock drifts are measured so that nodes know when to

The design aims to optimise energy conservation rather than system throughput, in many sensing scenarios high throughput is not required. Sensors collect and report some physical data such as temperature and humidity. In such cases, the data reporting rate may be in minutes or hours. Two packet structures are used in PoRAP. The control packet is used in control and setup phase. It contains essential information for transmission power adaptation and scheduling. The data packet is used to deliver the collected physical data back to the

The remaining parts of this chapter is organised as follows: Section 2 addressed application specific WSN. At present, WSN has been used in both military and civil applications. Each application category has particular characteristics and its own set of requirements. Hence, there are significant challenges in a generic protocol design for a variety of applications. Resource constraint issues are provided in Section 3. Apart from limited power resources, sensors also have constrained communication ranges for indoor and outdoor environments. The distance between the source and destination is crucial to employing an appropriate underlying communication paradigm. Section 4 describes the experimental details and their results which motivate the design of PoRAP. There are several factors which affect the link quality metrics such as distance between source and base station and time of day. The design of PoRAP is outlined in Section 5. PoRAP consists of several TinyOS components at the source and base station. The results shown in Section 4 motivates the design. The results of PoRAP evaluation in terms of energy conservation are presented in Section 6. Lower transmission power can be used to save the power whilst the reliability is in the desired

Apart from being used in military or surveillance, WSN has been deployed in several civil applications which have different requirements. Periodic sensing is required in some habitat and environmental monitoring systems whilst event sensing is the norm in surveillance systems. Network lifetime and data reporting rates are therefore major concerns for the

wake up to listen and transmit. Delays depend upon payload size.

range. Finally, Section 7 concludes the chapter.

**2. Application specific WSN** 

packet loss is avoided.

base station.

former and latter cases, respectively. To be application specific results in a more complicated design process, especially in the case of designing a generic power-aware protocol.

In total, seven groups of applications have been categorised by us based upon their functionalities including habitat monitoring (HM) (Juang et al., 2002; Mainwaring et al., 2002; Szewczyk et al., 2004), environmental monitoring (EM) (Allen et al., 2006; Martinez et al., 2005), health monitoring (HEM) (Jovanov et al., 2003, Otto et al., 2006), structural health monitoring (SHM) (Chintalapudi et al., 2006; Kottapalli et al., 2003; Paek et al., 2005, Schmid et al., 2005), event detection and tracking (EDT) (Arora et al., 2004; Dreicer et al., 2002; Simon et al., 2004), transport monitoring (TM) (Coleri et al., 2004) and location-aware system (LAS) (Brignone et al., 2005). Specific capabilities and underlying communication paradigms have been outlined. For example, data encryption may be required in some health monitoring systems for transmitting a patient's diagnosis data to the main server located at the hospital. Furthermore, data correctness is also required in this case. In some applications such as event tracking and detection systems, several intermediate nodes are required for forwarding the sensed data to the base station. However, a direct communication from source to base station is found in some health monitoring systems. This section addresses application specific characteristics of WSN applications by detailing the differences in their requirements.

#### **2.1 Event/periodic based**

The "Event/Periodic Based" aspect demonstrates how often data reporting is conducted. There are three main types including event-based, periodic-based and hybrid. Each sensor is triggered to operate by the occurrence of an event in the case of an event-based application. An example of this application type is the Event Detection and Tracking. Congestion is one of the major concerns designing a protocol to support event-based networking as it is caused by a lot of traffic generated by all sources in an event area. The key idea of congestion avoidance is to control data reporting rate of such sensors (Sankarasubramaniam et al., 2003). However, the main assumption is that all data packets have the same priority. Packet loss is therefore tolerantly acceptable. There are several works on congestion control specifically developed for WSN (Ee & Bajcsy, 2004; Hull et al., 2004, Lu et al., 2002, Wan et al., 2003). The congestion control approach focuses on channel monitoring to dynamically adjust the data forwarding rate. CODA (Wan et al., 2003) has been designed to cover two types of problems corresponding to the deployed sensors and their data rate. However, it does not provide any queue occupancy monitoring. Sending an ACK (Acknowledgement) in the case of persistent congestion, even if it is small in size, may increase the number of traffic. This mechanism also requires feedback signalling which results in higher cost. Only packet prioritisation could be found in (Lu et al., 2002). However, it proposes the VMS (Velocity Monotonic Scheduling) policy which supports both static and dynamic computation of the requested velocity and it also solves the fairness problem. Both channel and queue occupancy monitoring are provided in (Hull et al., 2004) and (Ee & Bajcsy, 2004). A child node can transmit packets only when its parent does not experience congestion problems and some help from the MAC (Medium Access Control) layer to shift the transmitting time to avoid interference are proposed in (Hull et al., 2004). A similar concept also exists in (Ee & Bajcsy, 2004) by comparing the normalised rate of a node and its parents.

Each sensor periodically performs its operation. Some examples of data collected by the sensors are temperature and humidity. The significant change in readings may be used to

of power, taken from the battery power available. Introducing several intelligent features to

Each source can transmit the data directly to the base station if the sources are located within the base station's communication range. Some examples of existing applications deploying single-hop communication (Mainwaring et al., 2002; Martinez et al., 2005; Jovanov et al., 2003; Otto et al., 2006). For single-hop, the sources are located within the base station's range. Direct communication is therefore feasible and several benefits are realised. One of the advantages is the ability to introduce a variety of intelligent features to the base station as it is assumed to have more power and computational capabilities compared to an ordinary sensor. Each source does not require the power necessary for routing. Idle listening can be minimised as the sources can be switched to sleep mode if they do not transmit data or receive the control packet. The base station controls the communication schedule of its sources to avoid data collisions. Power for carrier sensing is not desired. In multi-hop, each source is responsible for sensing, data reporting and routing. The number of transmissions and receptions increases with the number of intermediary nodes required for data

This work looks at protocol development for single-hop. A scenario where the single-hop is viable is Environmental Monitoring (EM). Sources and base stations are distributed and several clusters or patches are formed. A power-aware, single-hop protocol can thus be used in each of the clusters (Mainwaring et al., 2002). A low duty cycle is the norm in EM so the communication cycle of each source can be scheduled by the base station. A time slot is allocated to each source to perform data transmissions. Carrier sensing is thus not required in this scheme. The sources synchronise to the base station by checking the information

Wireless sensor network (WSN) has been currently deployed in several civil applications. The physical data is collected and transmitted for further analysis. The issue of reliability in data delivery is important for providing complete reliability consumes a significant proportion of power. Applying the Transmission Control Protocol (TCP) protocol to WSN is expensive because of its three-way handshake mechanism and packet header size. The User Datagram Protocol (UDP) is considered to be more suitable for sensors although it was designed to provide unreliable data transport. In some applications, data loss may be not a serious problem because of the large amount of deployed sensors. Reliable data transport is important for some types of data such as control messages delivered by the base station (Wan et al., 2002). The following paragraphs provide some details of reliable transport protocol for WSN researches including PSFQ (Pump Slowly, Fetch Quickly) (Wan et al., 2002), ESRT (Event-to-Sink Reliable Transport) (Sankarasubramaniam et al., 2003), and

One of the main goals to achieve reliable data transport is to orchestrate data receiving and forwarding processes to lessen the packet loss due to buffer overflow. PSFQ proposes three different operations including pump, fetch and report. Data generated from a source node is injected slowly into the network in order to allow such nodes experiencing data loss to fetch the missing packets very aggressively. Timing is a core process in order to avoid operational synchronisation. A hop-by-hop recovery is applied to avoid exponential error accumulation as occurs in the end-to-end scheme. Data delivery status information can be sent back to

RMST (Reliable Multi-Segment Transport) (Stann & Heidemann, 2003).

users or applications in a piggyback fashion.

each sensor is also limited due to the power constraint.

forwarding.

**2.4 Reliability** 

included in the control packet.

identify the presence of seabirds (Mainwaring et al., 2002) and intruders (Arora et al., 2004). Instead of heavily generated traffics, both sensor and network lifetimes are the core requirement of this application type. Finally, both event and periodic sensing operations may be desired in some applications such as SHM (Structural Health Monitoring) and EDT systems. For example, the displacement of construction elements is periodically reported for maintenance purposes whilst an event-based operation is applied for warning and evacuating notifications during an earthquake.

This work focuses on developing a power-aware protocol which supports an efficient data delivery in periodic based applications such as health, habitat and environmental monitoring where the data reporting rate is in minutes or hours. Sensors may be scattered over a remote and hostile area to collect and report physical data and they should have to operate for months. Hence, battery lifetime is important and one of the main goals is to conserve communication energy.

#### **2.2 Mobility of sources**

The mobility of sources or sensors can be found in some particular applications such as HM (Habitat Monitoring, HEM (HEalth Monitoring and LAS (Location-Aware System). In some cases, sensors are attached to the targeted objects or location (Jovanov et al., 2003; Juang et al., 2002, Martinez et al., 2005) in order to monitor the data of interest or current location. Mobile sensor networks have a different set of supporting infrastructures compared to the traditional WSN. It is essential for each mobile sensor to know its own location. The GPS (Global Positioning System) is used for locating sensors which are attached to the goods. Alternatively, several nodes with known locations may be used as references for the others to calculate their own locations [Brignone et al., 2005]. The issues of sensor mobility are beyond the scope of this work.

#### **2.3 Mobility of sources**

Wireless sensor network (WSN) consists of sensors which are wirelessly connected. The main objective of WSN development is to collect physical data from an area of interest. Therefore, communication between sensors is a key aspect. Normally there are two node types in WSN including the source and base station. Sources are ordinary sensors having limited resources whereas base stations are assumed to have more power and other resources. The main duty of sensors is collecting and transmitting data to the destination or base station. The sensors are probably required to cover a large area and direct communication between sources and base station is unlikely due to limited communication range. Several intermediate sensors responsible for forwarding data packets to the base station are therefore required. This is known as multi-hop communication. Each sensor also acts as a routing node in order to find the shortest or cheapest path by means of power consumption. Several applications deploy multi-hop communication (Allen et al., 2006; Chintalapudi et al., 2006; Schmid et al., 2005; Dreicer et al., 2002; Simon et al., 2004). The multi-hop approach has several advantages. For example, a new path is discovered when some sensors die. Deploying a large number of cheap sensors over a large area is feasible as the sensors can act as routing nodes and the collected data is forwarded to the destination. However, one of its drawbacks is each node has to listen to the channel most of the time in order to detect if a message is arriving. The sensors have to conduct some computations in order to discover the cheapest path. Moreover, communication with its neighbours is another requirement to set up a selected path. Such processes require a significant amount

identify the presence of seabirds (Mainwaring et al., 2002) and intruders (Arora et al., 2004). Instead of heavily generated traffics, both sensor and network lifetimes are the core requirement of this application type. Finally, both event and periodic sensing operations may be desired in some applications such as SHM (Structural Health Monitoring) and EDT systems. For example, the displacement of construction elements is periodically reported for maintenance purposes whilst an event-based operation is applied for warning and

This work focuses on developing a power-aware protocol which supports an efficient data delivery in periodic based applications such as health, habitat and environmental monitoring where the data reporting rate is in minutes or hours. Sensors may be scattered over a remote and hostile area to collect and report physical data and they should have to operate for months. Hence, battery lifetime is important and one of the main goals is to

The mobility of sources or sensors can be found in some particular applications such as HM (Habitat Monitoring, HEM (HEalth Monitoring and LAS (Location-Aware System). In some cases, sensors are attached to the targeted objects or location (Jovanov et al., 2003; Juang et al., 2002, Martinez et al., 2005) in order to monitor the data of interest or current location. Mobile sensor networks have a different set of supporting infrastructures compared to the traditional WSN. It is essential for each mobile sensor to know its own location. The GPS (Global Positioning System) is used for locating sensors which are attached to the goods. Alternatively, several nodes with known locations may be used as references for the others to calculate their own locations [Brignone et al., 2005]. The issues of sensor mobility are

Wireless sensor network (WSN) consists of sensors which are wirelessly connected. The main objective of WSN development is to collect physical data from an area of interest. Therefore, communication between sensors is a key aspect. Normally there are two node types in WSN including the source and base station. Sources are ordinary sensors having limited resources whereas base stations are assumed to have more power and other resources. The main duty of sensors is collecting and transmitting data to the destination or base station. The sensors are probably required to cover a large area and direct communication between sources and base station is unlikely due to limited communication range. Several intermediate sensors responsible for forwarding data packets to the base station are therefore required. This is known as multi-hop communication. Each sensor also acts as a routing node in order to find the shortest or cheapest path by means of power consumption. Several applications deploy multi-hop communication (Allen et al., 2006; Chintalapudi et al., 2006; Schmid et al., 2005; Dreicer et al., 2002; Simon et al., 2004). The multi-hop approach has several advantages. For example, a new path is discovered when some sensors die. Deploying a large number of cheap sensors over a large area is feasible as the sensors can act as routing nodes and the collected data is forwarded to the destination. However, one of its drawbacks is each node has to listen to the channel most of the time in order to detect if a message is arriving. The sensors have to conduct some computations in order to discover the cheapest path. Moreover, communication with its neighbours is another requirement to set up a selected path. Such processes require a significant amount

evacuating notifications during an earthquake.

conserve communication energy.

beyond the scope of this work.

**2.3 Mobility of sources** 

**2.2 Mobility of sources** 

of power, taken from the battery power available. Introducing several intelligent features to each sensor is also limited due to the power constraint.

Each source can transmit the data directly to the base station if the sources are located within the base station's communication range. Some examples of existing applications deploying single-hop communication (Mainwaring et al., 2002; Martinez et al., 2005; Jovanov et al., 2003; Otto et al., 2006). For single-hop, the sources are located within the base station's range. Direct communication is therefore feasible and several benefits are realised. One of the advantages is the ability to introduce a variety of intelligent features to the base station as it is assumed to have more power and computational capabilities compared to an ordinary sensor. Each source does not require the power necessary for routing. Idle listening can be minimised as the sources can be switched to sleep mode if they do not transmit data or receive the control packet. The base station controls the communication schedule of its sources to avoid data collisions. Power for carrier sensing is not desired. In multi-hop, each source is responsible for sensing, data reporting and routing. The number of transmissions and receptions increases with the number of intermediary nodes required for data forwarding.

This work looks at protocol development for single-hop. A scenario where the single-hop is viable is Environmental Monitoring (EM). Sources and base stations are distributed and several clusters or patches are formed. A power-aware, single-hop protocol can thus be used in each of the clusters (Mainwaring et al., 2002). A low duty cycle is the norm in EM so the communication cycle of each source can be scheduled by the base station. A time slot is allocated to each source to perform data transmissions. Carrier sensing is thus not required in this scheme. The sources synchronise to the base station by checking the information included in the control packet.

#### **2.4 Reliability**

Wireless sensor network (WSN) has been currently deployed in several civil applications. The physical data is collected and transmitted for further analysis. The issue of reliability in data delivery is important for providing complete reliability consumes a significant proportion of power. Applying the Transmission Control Protocol (TCP) protocol to WSN is expensive because of its three-way handshake mechanism and packet header size. The User Datagram Protocol (UDP) is considered to be more suitable for sensors although it was designed to provide unreliable data transport. In some applications, data loss may be not a serious problem because of the large amount of deployed sensors. Reliable data transport is important for some types of data such as control messages delivered by the base station (Wan et al., 2002). The following paragraphs provide some details of reliable transport protocol for WSN researches including PSFQ (Pump Slowly, Fetch Quickly) (Wan et al., 2002), ESRT (Event-to-Sink Reliable Transport) (Sankarasubramaniam et al., 2003), and RMST (Reliable Multi-Segment Transport) (Stann & Heidemann, 2003).

One of the main goals to achieve reliable data transport is to orchestrate data receiving and forwarding processes to lessen the packet loss due to buffer overflow. PSFQ proposes three different operations including pump, fetch and report. Data generated from a source node is injected slowly into the network in order to allow such nodes experiencing data loss to fetch the missing packets very aggressively. Timing is a core process in order to avoid operational synchronisation. A hop-by-hop recovery is applied to avoid exponential error accumulation as occurs in the end-to-end scheme. Data delivery status information can be sent back to users or applications in a piggyback fashion.

minimum and maximum transmission power is 8.5 and 17.4 milli-amperes (mA). Over 50%

Sensors equipped with CC2420 radio chips consume a greater amount of power when they receive data. According to the data sheet, 19.7mA is required for reception. Listening and sleeping consume 365 and 20 micro-amperes (µA), respectively. Hence, in the case of the CC2420 mote, data reception consumes more energy than transmission. The base station is the destination and it may be connected to a desktop or laptop computer. In such cases, the base station has extra power from the connected machine. However, the sensors which act as intermediary nodes between source and destination have to receive and forward packets resulting in sensor's lifetimes being decreased. The listening power is approximately 17 times greater than sleeping. In some applications such as environmental monitoring, the data sampling interval may be in minutes or hours. The transceivers should be switched to sleep mode instead of listening. Scheduling issues occur when two nodes communicate with each other. The data is not received if the receiver is in sleep mode. The nodes have to agree upon the same scheduling. Another scheme based upon contention-based can be used; the receiver can periodically listen to the signal propagated over the medium to inspect whether

WSN is also a shared medium system. Each of the sources and base station has to engage the medium to perform data communication. Data collisions occur if the sources transmit at the same time and energy will be wasted by unsuccessful data delivery. A Medium Access Control (MAC) protocol is required to resolve the contention. The features of the MAC protocol together with the application behaviour determine when a node is idle, when it is listening and when it is sending. As each of these states have different power requirements the MAC protocol impacts upon the efficiency of operation and the power consumption. There are two main MAC schemes; the contention and the schedule based. The medium is sensed prior to transmission and the sensors have to backoff if the medium is declared busy. This work focuses on the single-hop where the sources send data directly to the base station. Another scheme, schedule based, is adopted. A data slot is allocated to each node. No carrier sensing and corresponding energy is required. The main issue is that the slot must be long enough for completing data delivery, otherwise, data collisions are likely. Experimentations required to determine the duration required for both sending and receiving together with the effective factors such as data payload size. Each node is switched

to sleep mode to spend the least amount of power when its slot does not arrive.

minimised in order to decrease transmission and reception energy.

The buffering capacity of CC2420 is limited to 128 bytes. Taking the header's and footer's sizes into account, the allowable data payload size is thus less than 128 bytes. Apart from sensed data, some control information is required in the packet such as identification and timestamp. Additional packet structures may be required if all the information cannot be stored in one packet. Control overhead is considered as one of the costs and should be

Wireless sensor network (WSN) has been currently deployed in several surveillance and civil applications. Sensors may be scattered over an area of interest which can be very large. The communication range is thus important and depends upon the selected transceiver. For example, the CC2420 mote has 50m and 125m indoor and outdoor ranges. Under some circumstances, the maximum transmission power may not produce the maximum ranges. Furthermore, sending data to the node located at farther distances requires higher transmission power. Multi-hop is therefore usually used in WSN. Several intermediary sensors are required for data forwarding from the source to destination. Single-hop

of the power can be saved if the minimum power is always used.

the incoming message is destined for it.

Focusing only on the forward or sensor-to-sink direction, ESRT was designed to provide a reliable data transport by inspecting current network state in terms of reliability and congestion. The state result is categorised and the reporting frequency is then repetitively adjusted to reach an optimal point. ESRT provides both reliable data transport and congestion control. Local buffer level monitoring is used to detect congestion.

Directed Diffusion (Intanagonwiwat et al., 2003) is a routing protocol which provides a multipoint-to-multipoint communication. A sink firstly indicates an interest and propagates it to the nodes. Interest and node information is kept as gradients. The optimised reinforced path is then established to send the attribute-value pairs data. RMST is implemented as a filter to provide some information about the data fragment such as ID and total number of fragments to detect loss. A NACK (Negative ACKnowledgement) is sent via a back-channel to upstream neighbouring nodes in case of data loss.

According to the above fundamental protocol descriptions, several conclusions can be made. In a densely deployed environment, data loss may be accepted. However, this condition may apply only in the case of sensor-to-sink traffic. The sink or base station plays a major role in the network by broadcasting several control packets to the sensors. Such packets should not be lost. Moreover, there are various types of sensing data, such as structural displacement due to wind or earthquake (Xu et al., 2004), which need some combination from different nodes to create usable data before forwarding that data to the sink. PSFQ designing concepts are more complicated but can be applied to a broader area of application. The data retransmission mechanisms are not mentioned in ESRT as it focuses on statistical reliability. However, PSFQ does not provide congestion control schemes as ESRT does. RMST is designed to run over the Directed Diffusion routing protocol. Although it may take the least effort compared to the other two, it is not generic enough.

#### **3. Resource constraint issues**

This section introduces several issues of resource constraint in WSN. A sensor can be considered as a small computing device which is capable of sensing, data processing, storage and communication. Sensors are deployed in an area of interest and they may have to operate without maintenance throughout their lifetimes. Power is thus one of the limited resources. Unless an external source of energy is provided, power for all operations comes from batteries. Two AA batteries are required in the widely used platforms such as Tmote, Telos and Mica. The capacity of the AA battery is approximately 2,000 to 3,000 milli-amperehour (mAh). In order to calculate the battery life, the capacity is divided by the actual load current and the obtained lifetime is in hours. An equation for calculating sensor's lifetime is given in (Polastre et al., 2004) where the lifetime is equal to the product between capacity (mAh) and voltage (3V) divided by total energy consumption in micro-joules. The default capacity defined in (Polastre et al., 2004) is set at 2,500mAh.

Communication accounts for a significant proportion of energy consumption. There are four main modes of communication including sending, receiving, sleeping and listening. The transceiver is one of the major sensor components and it makes them capable of communicating with other nodes. Recent transceivers or radio chips such as CC1000 and CC2420 provide programmable transmission power. Sensors consume less power when they send at a lower power level. Hence, transmission power control is one of power-aware schemes in WSN. The sensors do not always send at the maximum power. Tmote platform is chosen in this study and it employs CC2420 transceiver. For the CC2420 mote the

Focusing only on the forward or sensor-to-sink direction, ESRT was designed to provide a reliable data transport by inspecting current network state in terms of reliability and congestion. The state result is categorised and the reporting frequency is then repetitively adjusted to reach an optimal point. ESRT provides both reliable data transport and

Directed Diffusion (Intanagonwiwat et al., 2003) is a routing protocol which provides a multipoint-to-multipoint communication. A sink firstly indicates an interest and propagates it to the nodes. Interest and node information is kept as gradients. The optimised reinforced path is then established to send the attribute-value pairs data. RMST is implemented as a filter to provide some information about the data fragment such as ID and total number of fragments to detect loss. A NACK (Negative ACKnowledgement) is sent via a back-channel

According to the above fundamental protocol descriptions, several conclusions can be made. In a densely deployed environment, data loss may be accepted. However, this condition may apply only in the case of sensor-to-sink traffic. The sink or base station plays a major role in the network by broadcasting several control packets to the sensors. Such packets should not be lost. Moreover, there are various types of sensing data, such as structural displacement due to wind or earthquake (Xu et al., 2004), which need some combination from different nodes to create usable data before forwarding that data to the sink. PSFQ designing concepts are more complicated but can be applied to a broader area of application. The data retransmission mechanisms are not mentioned in ESRT as it focuses on statistical reliability. However, PSFQ does not provide congestion control schemes as ESRT does. RMST is designed to run over the Directed Diffusion routing protocol. Although it

This section introduces several issues of resource constraint in WSN. A sensor can be considered as a small computing device which is capable of sensing, data processing, storage and communication. Sensors are deployed in an area of interest and they may have to operate without maintenance throughout their lifetimes. Power is thus one of the limited resources. Unless an external source of energy is provided, power for all operations comes from batteries. Two AA batteries are required in the widely used platforms such as Tmote, Telos and Mica. The capacity of the AA battery is approximately 2,000 to 3,000 milli-amperehour (mAh). In order to calculate the battery life, the capacity is divided by the actual load current and the obtained lifetime is in hours. An equation for calculating sensor's lifetime is given in (Polastre et al., 2004) where the lifetime is equal to the product between capacity (mAh) and voltage (3V) divided by total energy consumption in micro-joules. The default

Communication accounts for a significant proportion of energy consumption. There are four main modes of communication including sending, receiving, sleeping and listening. The transceiver is one of the major sensor components and it makes them capable of communicating with other nodes. Recent transceivers or radio chips such as CC1000 and CC2420 provide programmable transmission power. Sensors consume less power when they send at a lower power level. Hence, transmission power control is one of power-aware schemes in WSN. The sensors do not always send at the maximum power. Tmote platform is chosen in this study and it employs CC2420 transceiver. For the CC2420 mote the

congestion control. Local buffer level monitoring is used to detect congestion.

may take the least effort compared to the other two, it is not generic enough.

capacity defined in (Polastre et al., 2004) is set at 2,500mAh.

to upstream neighbouring nodes in case of data loss.

**3. Resource constraint issues** 

minimum and maximum transmission power is 8.5 and 17.4 milli-amperes (mA). Over 50% of the power can be saved if the minimum power is always used.

Sensors equipped with CC2420 radio chips consume a greater amount of power when they receive data. According to the data sheet, 19.7mA is required for reception. Listening and sleeping consume 365 and 20 micro-amperes (µA), respectively. Hence, in the case of the CC2420 mote, data reception consumes more energy than transmission. The base station is the destination and it may be connected to a desktop or laptop computer. In such cases, the base station has extra power from the connected machine. However, the sensors which act as intermediary nodes between source and destination have to receive and forward packets resulting in sensor's lifetimes being decreased. The listening power is approximately 17 times greater than sleeping. In some applications such as environmental monitoring, the data sampling interval may be in minutes or hours. The transceivers should be switched to sleep mode instead of listening. Scheduling issues occur when two nodes communicate with each other. The data is not received if the receiver is in sleep mode. The nodes have to agree upon the same scheduling. Another scheme based upon contention-based can be used; the receiver can periodically listen to the signal propagated over the medium to inspect whether the incoming message is destined for it.

WSN is also a shared medium system. Each of the sources and base station has to engage the medium to perform data communication. Data collisions occur if the sources transmit at the same time and energy will be wasted by unsuccessful data delivery. A Medium Access Control (MAC) protocol is required to resolve the contention. The features of the MAC protocol together with the application behaviour determine when a node is idle, when it is listening and when it is sending. As each of these states have different power requirements the MAC protocol impacts upon the efficiency of operation and the power consumption. There are two main MAC schemes; the contention and the schedule based. The medium is sensed prior to transmission and the sensors have to backoff if the medium is declared busy. This work focuses on the single-hop where the sources send data directly to the base station. Another scheme, schedule based, is adopted. A data slot is allocated to each node. No carrier sensing and corresponding energy is required. The main issue is that the slot must be long enough for completing data delivery, otherwise, data collisions are likely. Experimentations required to determine the duration required for both sending and receiving together with the effective factors such as data payload size. Each node is switched to sleep mode to spend the least amount of power when its slot does not arrive.

The buffering capacity of CC2420 is limited to 128 bytes. Taking the header's and footer's sizes into account, the allowable data payload size is thus less than 128 bytes. Apart from sensed data, some control information is required in the packet such as identification and timestamp. Additional packet structures may be required if all the information cannot be stored in one packet. Control overhead is considered as one of the costs and should be minimised in order to decrease transmission and reception energy.

Wireless sensor network (WSN) has been currently deployed in several surveillance and civil applications. Sensors may be scattered over an area of interest which can be very large. The communication range is thus important and depends upon the selected transceiver. For example, the CC2420 mote has 50m and 125m indoor and outdoor ranges. Under some circumstances, the maximum transmission power may not produce the maximum ranges. Furthermore, sending data to the node located at farther distances requires higher transmission power. Multi-hop is therefore usually used in WSN. Several intermediary sensors are required for data forwarding from the source to destination. Single-hop

used. The corresponding current consumption was measured by (Shnayder et al., 2004) and their results are shown in Table 1. A simulation duration of 60 seconds and a total of 30 runs were conducted at each power level. A higher current will be consumed if the sensor

> **Transmission Power (dBm) Required Current (mA)**  -20 5.21 -10 6.10 0 8.47 +6 13.77 +10 21.48

The results shown in Table 1 were used to compute the energy consumption required by each transmission power level. Fig. 1 shows error-bar plots of radio and total energy consumption at -20, -10, 0, +6 and +10 dBm. An analysis of power usage and conservation

According to Fig. 1, several observations can be made. Firstly, an increase in transmission power results in a higher energy consumption. Transmitting data at lower power uses less energy. For example, over 75% of energy can be conserved if the minimum power is used for transmission instead of the maximum. Secondly, the radio unit consumes a significant amount of energy. Up to 56% and 84% of energy are used by the radio if the sensor transmits at minimum and maximum power levels, respectively. The results are validated by the CC1000 data sheet which is the employed radio in Mica2. According to the CC1000 datasheet, the required current consumption for -20 and +10 dBm are 6.9 and 26.7 milli-amp (mA), respectively. Therefore, over 74% can be conserved and this is close to the 75% which

Fig. 1. Radio and total energy consumption at various transmission power levels

Table 1. Current consumption measured by Shnayder et al., 2004

with respect to the maximum power level is described in Table 2.

transmits at a higher power.

is obtained from PowerTOSSIM.

communication is feasible if the destination is located within the source's range. Multiple transmissions and receptions are not required if direct communication applies. However, the same transmission power cannot always be used as the link quality changes over time. The next section describes several sources of variability in radio frequency
