**4. The WSN layered protocol stack**

The WSN layered protocol stack consists of the Application layer, the Transport layer, the Data link layer and the Physical layer. This section will cover the role of each layer and study its power consumption. The section will survey the current literature and analyze it with respect to power consumption.

#### **4.1 Application layer**

8 Will-be-set-by-IN-TECH

the WSN is inoperable. For many applications various protocols for maximizing the lifetime of the WSN are adequate while many other applications require WSNs to remain in remote areas for several months or years without opportunities for manual power replenishment. Many research centers have developed models to efficiently harvest energy for power as for sensing previously mentioned. A WSN which can obtain its power requirements from its surrounding environment essentially has an infinite lifetime. Various approaches from mechanical vibrational energy harvesting to photon collection schemes are being considered

Fig. 4. A low power wireless sensor node system powered from energy scavengers or

Figure 4 is a schematic of a low power WSN system that uses energy scavengers. In Figure 4, the energy sources are labeled Vsolar, Vvibe and Vbat for the solar, mechanical vibration and battery, respectively. A mutiplexer switches between the unregulated energy sources. ADC denotes the Analog to digital converter, DSP denotes the Digital signal processors and RF

Fig. 3. Transceiver

in an effort to self-generate power needs.

harvesters and a battery. Guilar et al. (2006).

denotes Radio frequency.

The application layer is in charge of collecting and processing sensor readings (including the use of data aggregation), performing time synchronization, implementing a security protocol (as needed) etc... Each one of these tasks uses one or more hardware modules and each task results in power being consumed.

### **4.1.1 Information fusion**

Traditionally, sensor motes were designed to perform very little to no processing. They would sense the environment and send the sensing data to the base station for processing. This resulted in large amounts of packets being sent from the motes to the base station. In addition, in several sensor network applications, the motes are exposed to conditions (such as very high/low temperatures, pressure and noise) that might sabotage their measurements. It was then proposed to use information fusion (also referred to as data aggregation) techniques at the motes in order to decrease the network traffic, save energy, remove outlier data, make predictions about future measurements and in general obtain better information quality by combining data from multiple sources. Data aggregation requires some amount of processing to be carried out at the motes. Data fusion can be used at different layers of the WSN protocol stack. For example, it can be used at the application layer to process sensor readings as well as at the network layer to consolidate routing information. In the following, we survey and analyze the work that has been done on data aggregation and information fusion.

Information fusion can be categorized into three classes. Complementary, redundant, and cooperative. This classification depends on the particular application and the relationship between the motes that gather the data. In the case of complementary information, sources gather different types of data and information fusion is applied to obtain a more complete picture from data. In the case of redundant information, one or more sources gather the same type of data and information fusion is used to discard the outlier measurements and filter the data for accuracy, reliability and confidence. In cooperative information fusion, two sources gather information that is fused to produce information that better represents the reality. Information fusion is performed for different purposes. In the following, we present a classification of data fusion algorithms based on the purpose of the information fusion.

Information fusion techniques could be either centralized or distributed. Centralized techniques have a single point that controls the fusion process but are simple to implement. However, all the sensor motes send their data to the central point, which overwhelms the central data point and floods the network with messages. Distributed techniques on the other hand are more complex to implement but are more energy efficient because the information is exchanged locally, which reduces the number of messages exchanged in the network. Several methods have been proposed for information fusion including inference, estimation, aggregation, and compression.

in NTP, clients synchronize their clocks to the server and servers are synchronized to using external time sources (using a GPS). NTP is not suitable for WSNs for a number of reasons. First, NTP is centralized. 2. In WSNs, it is impossible to accurately estimate the roundtrip time. 3. GPS is too expensive to use or is not an option for most WSN applications (for

Power Considerations for Sensor Networks 139

Most time synchronization protocols are sender to receiver. The sender time stamps a packet and the receiver extract the time stamp and tries to extrapolate its clock drift compared to the sender (Romer, 2001), (Ganeriwal et al., 2003). However, the Reference Broadcast Scheme (RBS) (Elson et al., 2002) is different. It is a receiver-to-receiver synchronization protocol. In RBS, a sender broadcasts a beacon without any time information. The receivers then exchange Acknowledgement messages with the time they received the beacon. Receivers can then extrapolate their own clock drift relative to each other. RBS works with two receivers and is easily extended to more than two receivers. In addition, increasing the number of broadcasts increases the accuracy of the scheme. Note that in RBS, the uncertainty of access time is removed (since the sender is removed from the drift calculations) and since the propagation time is assumed to be negligible in WSNs, the only uncertainty factor and potential error

The Timing sync Protocol for Sensor Networks (TPSN) (Romer, 2001), (Ganeriwal et al., 2003) is a sender-receiver protocol. In TPSN, the sender sends a packet to a receiver, which uses the TPSN equation to extract its local clock drift compared to the sender. TPSN then uses a tree hierarchy to propagate the synchronization, it categorizes the nodes in the network into levels during the discovery phase. During the synchronization phase, the root node (level 0) synchronizes all level 1 nodes. After this first phase of synchronization, level 1 nodes synchronize level 2 nodes and so on, until synchronization has been propagated through the entire network. TPSN achieves better accuracy than RBS when using MAC layer time stamps because RBS is limited by the transmission range and would require more beacons in order to

In (Greunen & Rabaey, 2003), the authors claim that most sensor network applications do not require very precise synchronization. In fact, they claim that most applications only require synchronization in the order of a fraction of a second. The authors therefore propose a different approach where the required accuracy is taken as a constraint and then a synchronization algorithm with minimal complexity is devised so that the requested accuracy can be achieved. In this work, the synchronization is propagated in a centralized manner where a spanning tree is created and synchronization is conducted along the edges of the tree. Centralized approaches to time synchronization are not energy efficient and often result in depleting the energy reserves of the root node. The authors in (Maroti et al., 2004) propose the Flooding Time Synchronization Protocol (FTSP). FTSP uses periodic flooding of synchronization messages. This approach makes the algorithm de-centralized, scalable and topology independent. In FTSP, the synchronization root is elected dynamically and re-elected periodically. The root is responsible for keeping the global time of the network. In this work, the nodes form a dynamic mesh like structure to propagate the time synchronization throughout the network (unlike TPSN). This work saves on the energy required to create an initial spanning tree (as in TPSN) and is therefore more energy efficient than TPSN. In addition, this protocol is not topology dependent and can perform synchronization even when

example, indoor applications will not have access to GPS signal).

margin in this protocol is the receive time.

perform synchronization.

Inference methods consist of making a decision based on previous knowledge. Protocols that have been proposed for inference include Bayesian inference algorithms (Coue et al., 2002; Tsymbal et al., 2003), Dempster-Shafer inference algorithms (Dempster, 1968; Shafer, 1976) and fussy logic algorithms (Gupta et al., 2005) among others.

Estimation methods use probabilistic theory to estimate a state based on a sequence of measurements. Estimation algorithms include the maximum likelihood algorithm (Xiao et al., 2005), least square, Kalman filter and particle filter (Kalman, 1960).

Data aggregation methods are used to overcome implosion and overlap and compression is used to reduce the amount of data by exploiting spatial correlation among the motes. Techniques for compression include distributed source coding (Xiong et al., 2004) and coding by ordering (Petriovic et al., 2003).

The implementation of these algorithms comes at a cost involving hardware complexity, CPU time and energy.

#### **4.1.2 Time synchronization**

Time synchronization consists of synchronizing the local clocks of all the members of a distributed network. In WSNs, it consists of synchronizing the clocks of all the motes in the network. Time synchronization is essential for all networked systems and is a requirement in most WSN applications and protocols. Example applications include environmental monitoring and target tracking among others. In these applications, the order of events is usually important. For example, in target tracking, sensors need to continuously report the location of a moving target, which could be time sensitive. Example protocols that require time synchronization are some MAC protocols (Demirkol et al., 2006) (such as the ones based on TDMA, where each node is assigned a time slot) in addition to several routing and security protocols.

In this section, we review the time synchronization algorithms proposed in the literature and analyze their power saving properties. In addition to being energy efficient, time synchronizations schemes for WSN need to be accurate and scalable.

When a packet is sent from node A to node B, node A can append a time stamp to the packet. Node B can then extract the time stamp from the packet, add the time it took the packet to travel from node A to node B (transmission time) in order to estimate its local clock's drift from node A. The packet delay consists of send time, access time, propagation time and receive time. Send time is the time interval between when the node issues the send command until the node is ready to send the packet.

The medium access times is the duration from when the node is ready to send until the time when the transmission starts. This is the step that makes time synchronization such a difficult problem. It is not possible to accurately estimate this time. The propagation time is the time it takes the packet to reach to the destination, and the receive time is the time it takes to receive the frame.

The Network Time Protocol (NTP) described in (NTP, n.d.), is the protocol that synchronizes the clocks in wired networked systems by estimating the roundtrip time of packets. It is the standard used on the Internet. NTP maintains a universal time (UTC) across the network. NTP is not suitable for WSNs because of its centralized nature and prohibitive cost. In fact 10 Will-be-set-by-IN-TECH

Inference methods consist of making a decision based on previous knowledge. Protocols that have been proposed for inference include Bayesian inference algorithms (Coue et al., 2002; Tsymbal et al., 2003), Dempster-Shafer inference algorithms (Dempster, 1968; Shafer, 1976)

Estimation methods use probabilistic theory to estimate a state based on a sequence of measurements. Estimation algorithms include the maximum likelihood algorithm (Xiao et

Data aggregation methods are used to overcome implosion and overlap and compression is used to reduce the amount of data by exploiting spatial correlation among the motes. Techniques for compression include distributed source coding (Xiong et al., 2004) and coding

The implementation of these algorithms comes at a cost involving hardware complexity, CPU

Time synchronization consists of synchronizing the local clocks of all the members of a distributed network. In WSNs, it consists of synchronizing the clocks of all the motes in the network. Time synchronization is essential for all networked systems and is a requirement in most WSN applications and protocols. Example applications include environmental monitoring and target tracking among others. In these applications, the order of events is usually important. For example, in target tracking, sensors need to continuously report the location of a moving target, which could be time sensitive. Example protocols that require time synchronization are some MAC protocols (Demirkol et al., 2006) (such as the ones based on TDMA, where each node is assigned a time slot) in addition to several routing and security

In this section, we review the time synchronization algorithms proposed in the literature and analyze their power saving properties. In addition to being energy efficient, time

When a packet is sent from node A to node B, node A can append a time stamp to the packet. Node B can then extract the time stamp from the packet, add the time it took the packet to travel from node A to node B (transmission time) in order to estimate its local clock's drift from node A. The packet delay consists of send time, access time, propagation time and receive time. Send time is the time interval between when the node issues the send command until

The medium access times is the duration from when the node is ready to send until the time when the transmission starts. This is the step that makes time synchronization such a difficult problem. It is not possible to accurately estimate this time. The propagation time is the time it takes the packet to reach to the destination, and the receive time is the time it takes to receive

The Network Time Protocol (NTP) described in (NTP, n.d.), is the protocol that synchronizes the clocks in wired networked systems by estimating the roundtrip time of packets. It is the standard used on the Internet. NTP maintains a universal time (UTC) across the network. NTP is not suitable for WSNs because of its centralized nature and prohibitive cost. In fact

synchronizations schemes for WSN need to be accurate and scalable.

and fussy logic algorithms (Gupta et al., 2005) among others.

by ordering (Petriovic et al., 2003).

the node is ready to send the packet.

**4.1.2 Time synchronization**

time and energy.

protocols.

the frame.

al., 2005), least square, Kalman filter and particle filter (Kalman, 1960).

in NTP, clients synchronize their clocks to the server and servers are synchronized to using external time sources (using a GPS). NTP is not suitable for WSNs for a number of reasons. First, NTP is centralized. 2. In WSNs, it is impossible to accurately estimate the roundtrip time. 3. GPS is too expensive to use or is not an option for most WSN applications (for example, indoor applications will not have access to GPS signal).

Most time synchronization protocols are sender to receiver. The sender time stamps a packet and the receiver extract the time stamp and tries to extrapolate its clock drift compared to the sender (Romer, 2001), (Ganeriwal et al., 2003). However, the Reference Broadcast Scheme (RBS) (Elson et al., 2002) is different. It is a receiver-to-receiver synchronization protocol. In RBS, a sender broadcasts a beacon without any time information. The receivers then exchange Acknowledgement messages with the time they received the beacon. Receivers can then extrapolate their own clock drift relative to each other. RBS works with two receivers and is easily extended to more than two receivers. In addition, increasing the number of broadcasts increases the accuracy of the scheme. Note that in RBS, the uncertainty of access time is removed (since the sender is removed from the drift calculations) and since the propagation time is assumed to be negligible in WSNs, the only uncertainty factor and potential error margin in this protocol is the receive time.

The Timing sync Protocol for Sensor Networks (TPSN) (Romer, 2001), (Ganeriwal et al., 2003) is a sender-receiver protocol. In TPSN, the sender sends a packet to a receiver, which uses the TPSN equation to extract its local clock drift compared to the sender. TPSN then uses a tree hierarchy to propagate the synchronization, it categorizes the nodes in the network into levels during the discovery phase. During the synchronization phase, the root node (level 0) synchronizes all level 1 nodes. After this first phase of synchronization, level 1 nodes synchronize level 2 nodes and so on, until synchronization has been propagated through the entire network. TPSN achieves better accuracy than RBS when using MAC layer time stamps because RBS is limited by the transmission range and would require more beacons in order to perform synchronization.

In (Greunen & Rabaey, 2003), the authors claim that most sensor network applications do not require very precise synchronization. In fact, they claim that most applications only require synchronization in the order of a fraction of a second. The authors therefore propose a different approach where the required accuracy is taken as a constraint and then a synchronization algorithm with minimal complexity is devised so that the requested accuracy can be achieved. In this work, the synchronization is propagated in a centralized manner where a spanning tree is created and synchronization is conducted along the edges of the tree.

Centralized approaches to time synchronization are not energy efficient and often result in depleting the energy reserves of the root node. The authors in (Maroti et al., 2004) propose the Flooding Time Synchronization Protocol (FTSP). FTSP uses periodic flooding of synchronization messages. This approach makes the algorithm de-centralized, scalable and topology independent. In FTSP, the synchronization root is elected dynamically and re-elected periodically. The root is responsible for keeping the global time of the network. In this work, the nodes form a dynamic mesh like structure to propagate the time synchronization throughout the network (unlike TPSN). This work saves on the energy required to create an initial spanning tree (as in TPSN) and is therefore more energy efficient than TPSN. In addition, this protocol is not topology dependent and can perform synchronization even when

that not all WSN applications can have access to FM signals especially for the applications

Power Considerations for Sensor Networks 141

In summary, in order for a time synchronization protocol to be appropriate for a wide range of WSN application, it needs to accurately compute the clock drifts, be distributed, scalable, adapt to any topology and be able to propagate the synchronization instantaneously and

The transport layer is mainly used to communicate with external networks (such as the

The network layer is in charge of all routing functions. Routing is the function that is used the most in multi-hop WSNs. It is the routing algorithm that allows nodes that are more than a hop away to communicate with each other and form a connected network. Because routing is used extensively in most WSN applications, it is the function that should be the most power efficient. A variety of routing protocols have been proposed in the literature, some of which are designed to be 'power aware' and use the battery level or the network lifetime as a routing constraint. This Section reviews these works and studies the effect of clustering on power

Initially, research on routing algorithms focused on Mobile Ad hoc NETworks (MANETs). In these networks, the nodes were designed to be highly mobile, which resulted in the development of on-demand routing algorithms. These algorithms use flooding to compute routes (see the Reliable Ad-Hoc On-Demand Distance Vector Routing Protocol (RAODV) (Khurana et al., 2006), and the Ad hoc On-demand Multipath Distance Vector (Marina & Das, 2001) among others). The traditional flooding method consists of every node broadcasting the data to all its neighbors, the neighbors broadcasting the data to their neighbors etc... Ultimately, the sink will overhear the data. Flooding-based protocols suffer from several inefficiencies including overwhelming the network with unnecessary transmissions, excessive energy consumption, implosion, overlap, among others see (Heinzelman et al., 1999). Routing in MANETs is a tedious problem because of their dynamic nature. Adding power efficiency

In (Mleki et al., 2002), the authors propose a reactive Power-aware Source Routing (PSR) protocol for MANETs. This protocol was based on the Dynamic Source Routing protocol (RFC4728, n.d.). PSR computes the cost of routes while taking into consideration both transmission power and remaining battery power. In PSR, the source broadcasts a message and intermediate nodes compute the path cost and add it to the header of the broadcast message. The destination then adds the least cost path to the reply and sends it back to the source. This solution fits the needs of MANETs but because of its broadcast nature, it is not suitable for the more resource constrained sensor networks. Since most sensor network applications require static sensors and are more resource constrained than MANETs, the routing solutions that were developed for MANETs are not suitable for the low power sensor networks. As a result, the Routing Over Low power and Lossy networks (ROLL) group was created as part of IETF in 2008 (Watteyne & Richichi, 2010) to help develop a standardized

Internet) and is therefore rarely implemented in sensor motes.

to the equation renders the problem even more tedious.

routing solution for sensor networks.

deployed in remote areas.

without flooding the network.

**4.2 Transport layer**

**4.3 Network layer**

consumption.

the topology of the network changes. However, the synchronization error in FTSP can grow exponentially with the size of the network (Lenzen et al., 2009).

Similar to FTP, the Novel Algorithm for Time Synchronization (NATS) is a decentralized time synchronization protocol. Unlike FTSP, NATS is a receiver-sender protocol because the receiver requests synchronization from the sender. This reduces the amount of messages exchanged for the purpose of time synchronization and therefore, reduces the amount of energy consumed during synchronization. NATS was designed at DePauw University by Peter Terlep <sup>1</sup> , Steven Klaback <sup>2</sup> and Khadija Stewart. NATS does not need to meet any specific topology prerequisites, it can adjust to topology changes. It accomplishes the following: 1) it does not need a third party device that is within radio communication range of all motes, 2) it does not need any one mote to be within range of all motes, 3) it is scalable, 4) it allows for deep sleep between synchronization activities, 5) it handles receiver-side medium access control (MAC) buffer latency uncertainty, 6) it addresses the inability to acquire a real-time sender-side MAC timestamp, 7) and it uses a distributed energy efficient algorithm for multi-hop synchronization.

Pair-wise synchronization in NATS starts when the root node receives a sync request. The root then sends two consecutive packets to the requesting node, each containing a timestamp at the MAC layer. The receiving node uses these two packets along with its receive time stamp to extrapolate the propagation and channel access times. It uses that information to estimate its clock drift from the root node. Time synchronization is then propagated throughout the network in a distributed manner, similar to FTSP, by having each synched node act as a potential root node for synchronization. Experimental results show that NATS provides better synchronization accuracy than TPSN. In fact, using the Sun Spots platform, the Mean Sync Error for NATS was 1.74ms versus 2.63ms for TPSN.

The Gradient Time Synchronization Protocol (GTSP) is completely distributed (Sommer & Wattenhofer, 2009), where the nodes periodically broadcast synchronization beacons to their neighbors and agree on a common clock. It is proven that after multiple beacon exchanges, the clock of the nodes converges to a common value. This algorithm is completely distributed and nodes only exchange beacons locally. GTSP is proven to achieve better time synchronization accuracy as compared to tree-based methods.

In PulseSync (Lenzen et al., 2009), the root node floods a "pulse" through the network in a breadth-first search tree manner. The nodes receiving the pulse then compensate for the drift relative to the root node. The authors note that the flooding of the pulse needs to be scheduled in order to avoid collisions. This protocol is proven to be accurate when used in sensor network applications where the topology does not change. In fact, it is proven to outperform FTSP by a factor of 5 on mid-size networks.

The authors in (Li et al., 2011), propose a new direction in time synchronization where the Radio Data System (RDS) of FM radios is used to synchronize the nodes' clocks. In this work, each node is equipped with an FM receiver and programmed to receive the same RDS signal. The mote's clock then uses a calibration component to calibrate itself to the RDS clock. The drawbacks of this method stem from the fact that the FM interface is not power efficient and

<sup>1</sup> Peter Terlep is currenty a Ph.D student at Michigan University

<sup>2</sup> Steven Klaback is currently with Digital Knowledge

that not all WSN applications can have access to FM signals especially for the applications deployed in remote areas.

In summary, in order for a time synchronization protocol to be appropriate for a wide range of WSN application, it needs to accurately compute the clock drifts, be distributed, scalable, adapt to any topology and be able to propagate the synchronization instantaneously and without flooding the network.

#### **4.2 Transport layer**

12 Will-be-set-by-IN-TECH

the topology of the network changes. However, the synchronization error in FTSP can grow

Similar to FTP, the Novel Algorithm for Time Synchronization (NATS) is a decentralized time synchronization protocol. Unlike FTSP, NATS is a receiver-sender protocol because the receiver requests synchronization from the sender. This reduces the amount of messages exchanged for the purpose of time synchronization and therefore, reduces the amount of energy consumed during synchronization. NATS was designed at DePauw University by Peter Terlep <sup>1</sup> , Steven Klaback <sup>2</sup> and Khadija Stewart. NATS does not need to meet any specific topology prerequisites, it can adjust to topology changes. It accomplishes the following: 1) it does not need a third party device that is within radio communication range of all motes, 2) it does not need any one mote to be within range of all motes, 3) it is scalable, 4) it allows for deep sleep between synchronization activities, 5) it handles receiver-side medium access control (MAC) buffer latency uncertainty, 6) it addresses the inability to acquire a real-time sender-side MAC timestamp, 7) and it uses a distributed energy efficient algorithm

Pair-wise synchronization in NATS starts when the root node receives a sync request. The root then sends two consecutive packets to the requesting node, each containing a timestamp at the MAC layer. The receiving node uses these two packets along with its receive time stamp to extrapolate the propagation and channel access times. It uses that information to estimate its clock drift from the root node. Time synchronization is then propagated throughout the network in a distributed manner, similar to FTSP, by having each synched node act as a potential root node for synchronization. Experimental results show that NATS provides better synchronization accuracy than TPSN. In fact, using the Sun Spots platform, the Mean Sync

The Gradient Time Synchronization Protocol (GTSP) is completely distributed (Sommer & Wattenhofer, 2009), where the nodes periodically broadcast synchronization beacons to their neighbors and agree on a common clock. It is proven that after multiple beacon exchanges, the clock of the nodes converges to a common value. This algorithm is completely distributed and nodes only exchange beacons locally. GTSP is proven to achieve better time synchronization

In PulseSync (Lenzen et al., 2009), the root node floods a "pulse" through the network in a breadth-first search tree manner. The nodes receiving the pulse then compensate for the drift relative to the root node. The authors note that the flooding of the pulse needs to be scheduled in order to avoid collisions. This protocol is proven to be accurate when used in sensor network applications where the topology does not change. In fact, it is proven to

The authors in (Li et al., 2011), propose a new direction in time synchronization where the Radio Data System (RDS) of FM radios is used to synchronize the nodes' clocks. In this work, each node is equipped with an FM receiver and programmed to receive the same RDS signal. The mote's clock then uses a calibration component to calibrate itself to the RDS clock. The drawbacks of this method stem from the fact that the FM interface is not power efficient and

exponentially with the size of the network (Lenzen et al., 2009).

for multi-hop synchronization.

Error for NATS was 1.74ms versus 2.63ms for TPSN.

accuracy as compared to tree-based methods.

outperform FTSP by a factor of 5 on mid-size networks.

<sup>1</sup> Peter Terlep is currenty a Ph.D student at Michigan University

<sup>2</sup> Steven Klaback is currently with Digital Knowledge

The transport layer is mainly used to communicate with external networks (such as the Internet) and is therefore rarely implemented in sensor motes.

### **4.3 Network layer**

The network layer is in charge of all routing functions. Routing is the function that is used the most in multi-hop WSNs. It is the routing algorithm that allows nodes that are more than a hop away to communicate with each other and form a connected network. Because routing is used extensively in most WSN applications, it is the function that should be the most power efficient. A variety of routing protocols have been proposed in the literature, some of which are designed to be 'power aware' and use the battery level or the network lifetime as a routing constraint. This Section reviews these works and studies the effect of clustering on power consumption.

Initially, research on routing algorithms focused on Mobile Ad hoc NETworks (MANETs). In these networks, the nodes were designed to be highly mobile, which resulted in the development of on-demand routing algorithms. These algorithms use flooding to compute routes (see the Reliable Ad-Hoc On-Demand Distance Vector Routing Protocol (RAODV) (Khurana et al., 2006), and the Ad hoc On-demand Multipath Distance Vector (Marina & Das, 2001) among others). The traditional flooding method consists of every node broadcasting the data to all its neighbors, the neighbors broadcasting the data to their neighbors etc... Ultimately, the sink will overhear the data. Flooding-based protocols suffer from several inefficiencies including overwhelming the network with unnecessary transmissions, excessive energy consumption, implosion, overlap, among others see (Heinzelman et al., 1999). Routing in MANETs is a tedious problem because of their dynamic nature. Adding power efficiency to the equation renders the problem even more tedious.

In (Mleki et al., 2002), the authors propose a reactive Power-aware Source Routing (PSR) protocol for MANETs. This protocol was based on the Dynamic Source Routing protocol (RFC4728, n.d.). PSR computes the cost of routes while taking into consideration both transmission power and remaining battery power. In PSR, the source broadcasts a message and intermediate nodes compute the path cost and add it to the header of the broadcast message. The destination then adds the least cost path to the reply and sends it back to the source. This solution fits the needs of MANETs but because of its broadcast nature, it is not suitable for the more resource constrained sensor networks. Since most sensor network applications require static sensors and are more resource constrained than MANETs, the routing solutions that were developed for MANETs are not suitable for the low power sensor networks. As a result, the Routing Over Low power and Lossy networks (ROLL) group was created as part of IETF in 2008 (Watteyne & Richichi, 2010) to help develop a standardized routing solution for sensor networks.

Information via Negotiation (SPIN) (Kulik, 1999) was the first work to suggest eliminating redundant information to save energy. Later, a series of protocols that use directed diffusion were proposed (Intanagonwiwat et al., 2000), (Braginsky & Estin, 2002), (Schurgers &

Power Considerations for Sensor Networks 143

An important step in routing in wireless sensor networks was the creation of routing algorithms based on directed diffusion, the first introduction is described in (Intanagonwiwat et al., 2000). In directed diffusion, a node sends a query for some particular data (data here is identified using an attribute-value pair). As a result, data matching the query description is "drawn" towards the querying node. The data can be aggregated by intermediate nodes and all the communication is only neighbor-to-neighbor. These types of algorithms achieve significant energy savings over the traditional broadcast-based algorithms. Despite the energy saving properties of the directed diffusion algorithms, they are not suitable for all sensor network applications. Some sensor network applications require continuous data flow from the sensors to the sink, as a consequence, query based algorithms will not be suitable for such applications since the sink would need to continuously query each sensor for data (Akkaya &

An alternate way of relaying information in WSNs, other than flooding, is gossiping (Kyasanur et al., 2006). In gossiping, the source node selects a random neighbor and forwards the data to them. The process continues until the destination is reached or a maximum number of hops is achieved. Similar to flooding protocols, gossiping protocols also waste energy by sending messages by sensors that cover overlapping areas. In addition, gossiping algorithms

An improvement to the traditional gossiping protocols is the location-based protocols. In these protocols, location information is used to direct the routing in order to reduce the number of transmissions and therefore save energy. One such protocol is SPEED (He et al., 2002). This protocol uses a combination of feedback control and non-deterministic geographic

In (Li et al., 2001), the authors propose an energy saving routing scheme called the adaptive max-minzPmin scheme. This routing algorithm selects a route that maximizes the minimum residual energy as long as it consumes no more than zPmin energy (Pmin energy is he amount of energy consumed by the minimum-energy route). This algorithm also computes the minimum node lifetime of the network and adjusts its routing criterion accordingly. While this method is hard to implement (keeping track of the lifetime of the nodes in a central

Another family of protocols is the hierarchical routing protocols. The main purpose of creating a hierarchy within a sensor network is to achieve scalability, i.e., the network performance should decrease slowly in response to an increase in the network size. The main form of hierarchical routing in WSNs is clustering, which consists of organizing the nodes into clusters where each cluster has a cluster head. The cluster head is then in charge of performing data aggregation or forward the packets on to the next hop. This leads to a smaller amount of data

One of the first clustering protocols, LEACH is described in (Heinzelman et al., 2000). LEACH randomly rotates the head cluster in order to balance the energy consumption amongst the nodes in the cluster and uses data fusion in order to reduce the amount of data sent to the sink.

can suffer from excessive delays because the next hop node is selected randomly.

forwarding to provide real-time unicast, area-multicast and real-time area-anycast.

location), it is more practical for ad hoc networks than it is for sensor networks.

being transmitted to the sink, which intrinsically saves energy.

Srivastava, 2001), (Chu et al., 2002).

Younis, 2003).

In (Watteyne & Richichi, 2010), the authors define a set of criteria that routing protocols must possess for routing in low-power and lossy networks. These criteria consist of satisfactory performance in: 1. Routing state. 2. Loss response. 3. Control cost. 4. Link cost. 5. Node cost. The authors then conclude that none of the mature IETF protocols, that were developed for MANETs, fulfill those requirements. The protocols examined in this work are: OSPF (RFC2328, n.d.), IS-IS (RFC1142, n.d.), OLSR (RFC3626, n.d.), OLSRv2 (draft-ietf-manet-olsrv2-12, n.d.), TBRPF (RFC3684, n.d.), RIP (RFC2453, n.d.), AODV (RFC3561, n.d.), DYMO (draft-ietf-manet-dymo-mib-04, n.d.), DSR (RFC4728, n.d.), IPv6 Neighbor Discovery (RFC4861, n.d.) and MANET-NHDP (draft-ietf-manet-nhdp-15, n.d.). In (Watteyne & Richichi, 2010), the authors suggest that a new protocol specification document needs to be created for routing in low-power and lossy networks. The discussion in (Watteyne & Richichi, 2010) was limited to mature and well documented IETF protocols, in the remaining of this section, we examine "energy aware" routing protocols designed for wireless sensor networks that have not been included in this review.

Routing algorithms with energy considerations aim to either minimize the energy consumption of the networks as a whole or increase the lifetime of the network. Protocols that attempt to minimize the energy consumption of the network usually compute and use the shortest paths in the network. As a consequence, a few select motes are usually overused and their energy reserve is depleted earlier than the rest of the motes. This could result in the network becoming partitioned and could therefore end its useful lifetime prematurely.

Most applications of WSNs are deployed in remote areas and are scheduled to monitor the area for long periods of time. In this case, extending the useful lifetime of the network is of at most importance. The concept of 'lifetime of the network' is difficult to define in WSNs (Dietrick & Dressler, 2009). For practical purposes, we define the useful lifetime of the network as: 'The total amount of time that the network is able to do useful work'. If for example the purpose of the network is to record sensor readings from ten different areas for as long as possible, the useful (operational or functional) lifetime of the network will be the total amount of time that at least one sensor is functional in each of the ten different areas and that there exists a path between each of those senors to the sink, i.e., those sensor motes are able to relay their readings to the sink. The useful lifetime of the network is therefore application specific and a uniform definition may not apply to all types of WSN applications.

The shortcomings of the broadcast-based protocols have led to the design of data-centric routing mechanisms. One of the earliest works on this type of protocols is SPINS (Heinzelman et al., 1999) where the data is named using high-level descriptors (meta-data). In this case, sensors exchange meta-data. The protocol relies on three types of messages: 1. ADV message, which is used to advertise particular meta-data, 2. REQ message used to request specific data, and 2. DATA message used to deliver the actual data. Spins achieves significant energy savings over traditional broadcast-based protocols (a factor of 3.5) and reduces the data redundancy in half. However, Spins does not guarantee the delivery of data to the requesting node, which makes this protocol unpractical for several applications of WSNs (Akkaya & Younis, 2003).

In data-centric routing algorithms, regions of sensors are queried to send their sensed readings to the sink. Because of the redundancy in sensors in each region, the data needs to be aggregated before it is forwarded to the sink. Several algorithms have been proposed to perform data aggregation to disregard the redundant information. Sensor Protocols for 14 Will-be-set-by-IN-TECH

In (Watteyne & Richichi, 2010), the authors define a set of criteria that routing protocols must possess for routing in low-power and lossy networks. These criteria consist of satisfactory performance in: 1. Routing state. 2. Loss response. 3. Control cost. 4. Link cost. 5. Node cost. The authors then conclude that none of the mature IETF protocols, that were developed for MANETs, fulfill those requirements. The protocols examined in this work are: OSPF (RFC2328, n.d.), IS-IS (RFC1142, n.d.), OLSR (RFC3626, n.d.), OLSRv2 (draft-ietf-manet-olsrv2-12, n.d.), TBRPF (RFC3684, n.d.), RIP (RFC2453, n.d.), AODV (RFC3561, n.d.), DYMO (draft-ietf-manet-dymo-mib-04, n.d.), DSR (RFC4728, n.d.), IPv6 Neighbor Discovery (RFC4861, n.d.) and MANET-NHDP (draft-ietf-manet-nhdp-15, n.d.). In (Watteyne & Richichi, 2010), the authors suggest that a new protocol specification document needs to be created for routing in low-power and lossy networks. The discussion in (Watteyne & Richichi, 2010) was limited to mature and well documented IETF protocols, in the remaining of this section, we examine "energy aware" routing protocols designed for

Routing algorithms with energy considerations aim to either minimize the energy consumption of the networks as a whole or increase the lifetime of the network. Protocols that attempt to minimize the energy consumption of the network usually compute and use the shortest paths in the network. As a consequence, a few select motes are usually overused and their energy reserve is depleted earlier than the rest of the motes. This could result in the network becoming partitioned and could therefore end its useful lifetime prematurely.

Most applications of WSNs are deployed in remote areas and are scheduled to monitor the area for long periods of time. In this case, extending the useful lifetime of the network is of at most importance. The concept of 'lifetime of the network' is difficult to define in WSNs (Dietrick & Dressler, 2009). For practical purposes, we define the useful lifetime of the network as: 'The total amount of time that the network is able to do useful work'. If for example the purpose of the network is to record sensor readings from ten different areas for as long as possible, the useful (operational or functional) lifetime of the network will be the total amount of time that at least one sensor is functional in each of the ten different areas and that there exists a path between each of those senors to the sink, i.e., those sensor motes are able to relay their readings to the sink. The useful lifetime of the network is therefore application specific

The shortcomings of the broadcast-based protocols have led to the design of data-centric routing mechanisms. One of the earliest works on this type of protocols is SPINS (Heinzelman et al., 1999) where the data is named using high-level descriptors (meta-data). In this case, sensors exchange meta-data. The protocol relies on three types of messages: 1. ADV message, which is used to advertise particular meta-data, 2. REQ message used to request specific data, and 2. DATA message used to deliver the actual data. Spins achieves significant energy savings over traditional broadcast-based protocols (a factor of 3.5) and reduces the data redundancy in half. However, Spins does not guarantee the delivery of data to the requesting node, which makes this protocol unpractical for several applications of WSNs (Akkaya &

In data-centric routing algorithms, regions of sensors are queried to send their sensed readings to the sink. Because of the redundancy in sensors in each region, the data needs to be aggregated before it is forwarded to the sink. Several algorithms have been proposed to perform data aggregation to disregard the redundant information. Sensor Protocols for

wireless sensor networks that have not been included in this review.

and a uniform definition may not apply to all types of WSN applications.

Younis, 2003).

Information via Negotiation (SPIN) (Kulik, 1999) was the first work to suggest eliminating redundant information to save energy. Later, a series of protocols that use directed diffusion were proposed (Intanagonwiwat et al., 2000), (Braginsky & Estin, 2002), (Schurgers & Srivastava, 2001), (Chu et al., 2002).

An important step in routing in wireless sensor networks was the creation of routing algorithms based on directed diffusion, the first introduction is described in (Intanagonwiwat et al., 2000). In directed diffusion, a node sends a query for some particular data (data here is identified using an attribute-value pair). As a result, data matching the query description is "drawn" towards the querying node. The data can be aggregated by intermediate nodes and all the communication is only neighbor-to-neighbor. These types of algorithms achieve significant energy savings over the traditional broadcast-based algorithms. Despite the energy saving properties of the directed diffusion algorithms, they are not suitable for all sensor network applications. Some sensor network applications require continuous data flow from the sensors to the sink, as a consequence, query based algorithms will not be suitable for such applications since the sink would need to continuously query each sensor for data (Akkaya & Younis, 2003).

An alternate way of relaying information in WSNs, other than flooding, is gossiping (Kyasanur et al., 2006). In gossiping, the source node selects a random neighbor and forwards the data to them. The process continues until the destination is reached or a maximum number of hops is achieved. Similar to flooding protocols, gossiping protocols also waste energy by sending messages by sensors that cover overlapping areas. In addition, gossiping algorithms can suffer from excessive delays because the next hop node is selected randomly.

An improvement to the traditional gossiping protocols is the location-based protocols. In these protocols, location information is used to direct the routing in order to reduce the number of transmissions and therefore save energy. One such protocol is SPEED (He et al., 2002). This protocol uses a combination of feedback control and non-deterministic geographic forwarding to provide real-time unicast, area-multicast and real-time area-anycast.

In (Li et al., 2001), the authors propose an energy saving routing scheme called the adaptive max-minzPmin scheme. This routing algorithm selects a route that maximizes the minimum residual energy as long as it consumes no more than zPmin energy (Pmin energy is he amount of energy consumed by the minimum-energy route). This algorithm also computes the minimum node lifetime of the network and adjusts its routing criterion accordingly. While this method is hard to implement (keeping track of the lifetime of the nodes in a central location), it is more practical for ad hoc networks than it is for sensor networks.

Another family of protocols is the hierarchical routing protocols. The main purpose of creating a hierarchy within a sensor network is to achieve scalability, i.e., the network performance should decrease slowly in response to an increase in the network size. The main form of hierarchical routing in WSNs is clustering, which consists of organizing the nodes into clusters where each cluster has a cluster head. The cluster head is then in charge of performing data aggregation or forward the packets on to the next hop. This leads to a smaller amount of data being transmitted to the sink, which intrinsically saves energy.

One of the first clustering protocols, LEACH is described in (Heinzelman et al., 2000). LEACH randomly rotates the head cluster in order to balance the energy consumption amongst the nodes in the cluster and uses data fusion in order to reduce the amount of data sent to the sink.

prevention mechanisms. In the rest of this Section, we study the main MAC layer protocols

Power Considerations for Sensor Networks 145

In this work, we consider energy efficiency to be the most important attribute in a MAC protocol. Other important attributes for a MAC protocol consist of providing fair and efficient

Most ad hoc network and WSN applications require the network to be deployed for an extended period of time. During their deployment, the motes are programmed to sense the environment and relay sensor readings to the sink. Several MAC protocols have been proposed for these applications where the motes are periodically scheduled to be in a power-saving state (a sleep state or an off state) in order to save their battery power and extend their deployment lifetime, see (Singh & Raghavendra, 1998; Stewart & Tragoudas, 2007; Ye et

PAMAS (Singh & Raghavendra, 1998) is a MAC protocol based on RTS/CTS. PAMAS schedules sleep intervals for sensor nodes to avoid overhearing and uses separate channels for data and control frames. In PAMAS, nodes probe their neighbors for transmission time in order to avoid collision as well. PAMAS reduces energy consumption by avoiding collision and transmission overhearing at the expense of increased hardware complexity, which in turn

The S-MAC (Ye et al., 2004) protocol reduces the energy consumption of the nodes by implementing the following mechanisms. First, it reduces idle listening by scheduling sleep intervals for nodes, in fact, S-MAC coordinates sleep intervals amongst neighboring nodes. Second, it divides long messages into smaller packets and transmits them back to back. As a result, nodes with longer messages occupy the wireless medium for longer periods of time. The authors show that this seemingly "unfair" advantage results in energy savings over traditional "fair" methods. Third, it implements a low-duty-cycle that reduces idle listening. Finally, it uses in-channel signaling to reduce overhearing by extending the work

S-MAC's mechanisms do reduce energy consumption at the expense of increased message latency. However, the predefined sleep intervals limit the flexibility of the protocol and the broadcast mechanism increases the probability for collision because S-MAC does not use

TMAC (Van & Langendoen, 2003) is similar to SMAC except that each node is equipped with a timer. In TMAC, a node is put on the low-power/sleep state if it does not transmit or receive for the entire duration of the timeout period. TMAC performs significantly better than S-MAC

In WiseMAC (El-Hoiydi & Decotignie, 2004), the authors propose a downlink (to be used when the sink transmits packets to sensors). WiseMAC uses non-persistent CSMA (np-CSMA) with preamble sampling in order to decrease idle listening. In this case, a preamble is used to alert the receiving node that a data packet is on its way. The preamble precedes each data packets. All the nodes in the medium listen to the medium for a constant time interval referred to as the sampling period. If a node hears a transmission while it is listening to the medium, it will continue to listen until it receives a frame or until the medium becomes idle. The sink precedes each data frame with a preamble sequence that is equal to the sampling period. This guarantees that the receiving node will be able to detect the transmission. On the downside,

that are proposed in the literature and analyze their power-saving properties.

access to the medium, scalability and adaptability to change.

al., 2004) among others.

affects the power consumption.

from PAMAS (Singh & Raghavendra, 1998).

RTS/CTS (Demirkol et al., 2006).

under variable load.

As a result, LEACH achieves significant energy savings compared to conventional routing protocols. Several other hierarchical protocols have been proposed in the literature who build up on LEACH such as TL-LEACH (Loscri et al., 2005) which proposes a two-level hierarchy to LEACH, EECS (Ye et al., 2006) where nodes compete for the position of cluster head, HEED (Yonis & Fahmy, 2004) where cluster heads are selected based on the distance between nodes, among others.

In (Iwanicki & Steen, 2009), the authors develop a framework to test the various hierarchical routing protocols proposed for WSNs. The authors state that hierarchical routing is a promising solution for the resource constrained WSNs and caution that the theoretical results presented in most hierarchical work can be very different from the results obtained using a more realistic framework. The proposed framework dismisses the idea of rotating the cluster head to save energy because this change complicates route computation by changing the routing addresses. The authors conclude that there is no one optimal hierarchical routing protocol for all WSN applications, rather protocols are application and requirement dependent.

In conclusion, there still exists the need to develop a low-frills, low-power, manageable and adaptable protocol for routing in the resource constrained sensor networks. The ROLL working group is still working on a requirement specification document. They may in fact, not be able to propose a single protocol for all or most WSN applications and could end up proposing or extending more than one protocol.

#### **4.4 Medium access control layer**

The main duties of sensor motes are communication, sensing and computing. Amongst these three tasks, communication consumes the most energy. It is therefore imperative to make sure that the communication task is as efficient as possible in order to prolong the energy lifetime of the motes. It is the data link layer that is responsible for establishing communication links between the motes, allowing the motes to share the wireless medium fairly and detecting/correcting transmission errors. Power considerations at the data link layer involve studying the hardware of the communication module (see Section 3) , the implementation of protocols such as the power management protocol and manipulating the power level of the transceiver.

The most energy waste occurs when a mote receives multiple frames at the same time. In this case all the frames that collide need to be discarded which results in wasted transmissions and receptions and increased latency. Other causes of energy waste are control packet overhead, overhearing unnecessary traffic and the long idle time in WSNs. In fact, in WSNs, idle listening consumes more than half the amount of energy required for reception (Ye et al., 2004). The Medium Access Control (MAC) layer is the sublayer of the data link layer that is responsible for handling the contention over the medium (in this case, the wireless medium). The main media access protocols used in wireless networks are Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Carrier Sense Multiple Access (CSMA), Request To Send/Clear To Send (RTS/CTS) protocols, and the IEEE 802.11 protocol. The purpose of these schemes is to avoid channel contention. In the following, we review the most relevant MAC protocols that are proposed for use with wireless sensor networks. The channel contention scheme in these protocols is based on the above described contention 16 Will-be-set-by-IN-TECH

As a result, LEACH achieves significant energy savings compared to conventional routing protocols. Several other hierarchical protocols have been proposed in the literature who build up on LEACH such as TL-LEACH (Loscri et al., 2005) which proposes a two-level hierarchy to LEACH, EECS (Ye et al., 2006) where nodes compete for the position of cluster head, HEED (Yonis & Fahmy, 2004) where cluster heads are selected based on the distance between nodes,

In (Iwanicki & Steen, 2009), the authors develop a framework to test the various hierarchical routing protocols proposed for WSNs. The authors state that hierarchical routing is a promising solution for the resource constrained WSNs and caution that the theoretical results presented in most hierarchical work can be very different from the results obtained using a more realistic framework. The proposed framework dismisses the idea of rotating the cluster head to save energy because this change complicates route computation by changing the routing addresses. The authors conclude that there is no one optimal hierarchical routing protocol for all WSN applications, rather protocols are application and requirement

In conclusion, there still exists the need to develop a low-frills, low-power, manageable and adaptable protocol for routing in the resource constrained sensor networks. The ROLL working group is still working on a requirement specification document. They may in fact, not be able to propose a single protocol for all or most WSN applications and could end up

The main duties of sensor motes are communication, sensing and computing. Amongst these three tasks, communication consumes the most energy. It is therefore imperative to make sure that the communication task is as efficient as possible in order to prolong the energy lifetime of the motes. It is the data link layer that is responsible for establishing communication links between the motes, allowing the motes to share the wireless medium fairly and detecting/correcting transmission errors. Power considerations at the data link layer involve studying the hardware of the communication module (see Section 3) , the implementation of protocols such as the power management protocol and manipulating the

The most energy waste occurs when a mote receives multiple frames at the same time. In this case all the frames that collide need to be discarded which results in wasted transmissions and receptions and increased latency. Other causes of energy waste are control packet overhead, overhearing unnecessary traffic and the long idle time in WSNs. In fact, in WSNs, idle listening consumes more than half the amount of energy required for reception (Ye et al., 2004). The Medium Access Control (MAC) layer is the sublayer of the data link layer that is responsible for handling the contention over the medium (in this case, the wireless medium). The main media access protocols used in wireless networks are Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Carrier Sense Multiple Access (CSMA), Request To Send/Clear To Send (RTS/CTS) protocols, and the IEEE 802.11 protocol. The purpose of these schemes is to avoid channel contention. In the following, we review the most relevant MAC protocols that are proposed for use with wireless sensor networks. The channel contention scheme in these protocols is based on the above described contention

among others.

dependent.

proposing or extending more than one protocol.

**4.4 Medium access control layer**

power level of the transceiver.

prevention mechanisms. In the rest of this Section, we study the main MAC layer protocols that are proposed in the literature and analyze their power-saving properties.

In this work, we consider energy efficiency to be the most important attribute in a MAC protocol. Other important attributes for a MAC protocol consist of providing fair and efficient access to the medium, scalability and adaptability to change.

Most ad hoc network and WSN applications require the network to be deployed for an extended period of time. During their deployment, the motes are programmed to sense the environment and relay sensor readings to the sink. Several MAC protocols have been proposed for these applications where the motes are periodically scheduled to be in a power-saving state (a sleep state or an off state) in order to save their battery power and extend their deployment lifetime, see (Singh & Raghavendra, 1998; Stewart & Tragoudas, 2007; Ye et al., 2004) among others.

PAMAS (Singh & Raghavendra, 1998) is a MAC protocol based on RTS/CTS. PAMAS schedules sleep intervals for sensor nodes to avoid overhearing and uses separate channels for data and control frames. In PAMAS, nodes probe their neighbors for transmission time in order to avoid collision as well. PAMAS reduces energy consumption by avoiding collision and transmission overhearing at the expense of increased hardware complexity, which in turn affects the power consumption.

The S-MAC (Ye et al., 2004) protocol reduces the energy consumption of the nodes by implementing the following mechanisms. First, it reduces idle listening by scheduling sleep intervals for nodes, in fact, S-MAC coordinates sleep intervals amongst neighboring nodes. Second, it divides long messages into smaller packets and transmits them back to back. As a result, nodes with longer messages occupy the wireless medium for longer periods of time. The authors show that this seemingly "unfair" advantage results in energy savings over traditional "fair" methods. Third, it implements a low-duty-cycle that reduces idle listening. Finally, it uses in-channel signaling to reduce overhearing by extending the work from PAMAS (Singh & Raghavendra, 1998).

S-MAC's mechanisms do reduce energy consumption at the expense of increased message latency. However, the predefined sleep intervals limit the flexibility of the protocol and the broadcast mechanism increases the probability for collision because S-MAC does not use RTS/CTS (Demirkol et al., 2006).

TMAC (Van & Langendoen, 2003) is similar to SMAC except that each node is equipped with a timer. In TMAC, a node is put on the low-power/sleep state if it does not transmit or receive for the entire duration of the timeout period. TMAC performs significantly better than S-MAC under variable load.

In WiseMAC (El-Hoiydi & Decotignie, 2004), the authors propose a downlink (to be used when the sink transmits packets to sensors). WiseMAC uses non-persistent CSMA (np-CSMA) with preamble sampling in order to decrease idle listening. In this case, a preamble is used to alert the receiving node that a data packet is on its way. The preamble precedes each data packets. All the nodes in the medium listen to the medium for a constant time interval referred to as the sampling period. If a node hears a transmission while it is listening to the medium, it will continue to listen until it receives a frame or until the medium becomes idle. The sink precedes each data frame with a preamble sequence that is equal to the sampling period. This guarantees that the receiving node will be able to detect the transmission. On the downside,

is charged with designing a physical layer, propagation effects due to the ambient conditions

Power Considerations for Sensor Networks 147

This chapter reviews the hardware architecture of wireless sensor motes, as well as their protocol stack focusing on power considerations at every level. We conclude that because of the diversity in WSN applications, it is very difficult to derive a universal power efficient

As far as the hardware components in WSNs, many advances have been made over the last few years. These improvements include more efficient apertures with better directivity and lower VSWR. The sensor element has been made to become more resolute while power management has improved due to the accessibility of more exotic materials for energy storage. The future holds near perfect antenna with nearly a 1:1 VSWR ensuring most of the energy leaving the system goes where it's designed to propagate. Researchers at Purdue University are working toward ensuring optical sensors are near perfectly efficient with

In terms of the WSN protocol stack, no one protocol has been adopted as a WSN standard, rather each protocol is designed to efficiently serve one or more WSN applications. The power efficiency of protocols has become the number one constraint in almost every layer of the protocol stack. More work is needed to design and develop protocols that are less application

Kemal Akkaya and Mohamed Younis, *A Survey on Routing Protocols for Wireless Sensor Networks*. Elsevier journal of Ad Hoc Networks, Volume 3, pages 325-349. I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, *Wireless sensor networks: a survey*.

S. Basagni, M. Conti, S. Giordano, Iv. Stojmenovic, *Chapter 11: Energy-efficient Communication in ad hoc Wireless Networks*. Mobile ad hoc networking, Wiley-IEEE Press, 2004. Bontempi, G., and Le Borgne, Y. (2005). *An adaptive modular approach to the mining of sensor*

D. Braginsky and D. Estin, *Rumor routing algorithm for sensor networks*. Proceedings of the

Wenshan Cai, and Vladmir Shalaev, *Optical Metamatrials: Fundamentals and Applications*.

M. Chu, H. Haussecker, and F. Zhao, *Scalable Information-Driven Sensor Querying and*

C. Coue, T. Franichard, P. Bessiere, and E. Mazer, *Multi-sensor data fusion using Bayesian*

Performance Computing Applications, Vol. 16, No. 3, August 2002.

*network data*. In Workshop on Data Mining in Sensor Networks, SIAM SDM, Newport

first workshop on Sensor Networks and Applications (WSNA), Atlanta, GA, October

*Routing for ad hoc Heterogeneous Sensor Networks*. The International Journal of High

*programming: An automotive application*. Proceedings of the IEEE/RSJ International Coference on Intelligent Robots and Systems. Vol. 1, Lausanne, Switzerland, pages

Elsevier Computer Networks, Volume 38, pages 393-422, 2002.

must be considered.

**5. Conclusion and future work**

specific and still power efficient.

ALERT, www.alertsystems.org

2002.

141-146.

Springer, 2010.

Beach, CA, USA, April.

**6. References**

architecture both in terms of hardware and software.

negative refractive metamaterials and photon collection efforts.

the long preamble sequence results in a low throughput and in increase power consumption. In addition, all the nodes within wireless range of the receiving node are able to hear the transmission. WiseMAC proposes an improvement to this where the sink takes advantage of knowing the sampling schedule of the nodes. The sink therefore, sends a smaller preamble and a frame right when the receiving node is scheduled to start sampling the medium.

WiseMAC suffers from two main drawbacks (Demirkol et al., 2006). The first drawback results from its decentralized sleep schedule where nodes wake up from their sleep cycle at different times. This is inefficient when broadcast communication is used because the broadcasted frames would need to be stored at the neighbors who are awake and end up being transmitted multiple times. The second drawback of the protocol is the fact that it is vulnerable to the hidden terminal problem where collision can happen at a node if it receives transmissions from two nodes that are not within transmission range of each other (Note that this is not a problem if WiseMAC is only used as a downlink protocol)

TRAMA (Rajendran et al., 2003) is a collision-free TDMA based MAC protocol for sensor networks. TRAMA ensures energy efficiency by avoiding collision during unicast, multicast and broadcast transmissions. In addition, in TRAMA, nodes can switch to a low-power state whenever they are not transmitting or receiving frames to save energy. In TRAMA, a node is elected to transmit within a two-hop neighborhood during each time slot. This mechanism avoids the hidden terminal problem.

TRAMA achieves significant energy savings due to: 1. the increased amount of low-power states, 2. the decreased amount of communication since the receiving nodes are indicated a priori, and 3. the significant decrease in collision probability. However, the latency when using TRAMA is longer compared to CSMA as a result of the high percentage of sleep time (Demirkol et al., 2006).

Berkeley MAC (B-MAC) (Polastre et al., 2004) is a low frills protocol based on clear channel assessment, it uses low power listening with preamble sampling. The default mode in B-MAC does not include a mechanism to avoid the hidden terminal problem, which could be implemented by higher layers if needed. B-MAC achieves significant energy savings when varying check time, by making the preamble constant and setting the sample rate. However, since the protocol is bare-bone, additional features would have to be implemented at higher layers when needed, which increases the complexity of the system as a whole.

Even though multiple MAC layer protocols provide adequate performance, no single protocol has been chosen as a standard. This is due to the fact that some protocols perform better than others for particular applications, communication pattern, network infrastructures and or network densities. An ideal energy efficient MAC layer protocol for WSNs would use a local schedule for motes to turn to the low-power/off state as a function of their residual energy as well as their sensing schedule. The schedule should aim to maximize the sleep time of the motes while preserving their sensing schedule, local connectivity and balancing their energy levels in order to increase the lifetime of the network as a whole.

#### **4.4.1 Physical layer**

Frequency detection, generation, modulation and coupling are the responsibility of the physical layer and are explained in detail in the hardware section. Note that when an engineer 18 Will-be-set-by-IN-TECH

the long preamble sequence results in a low throughput and in increase power consumption. In addition, all the nodes within wireless range of the receiving node are able to hear the transmission. WiseMAC proposes an improvement to this where the sink takes advantage of knowing the sampling schedule of the nodes. The sink therefore, sends a smaller preamble and a frame right when the receiving node is scheduled to start sampling the medium.

WiseMAC suffers from two main drawbacks (Demirkol et al., 2006). The first drawback results from its decentralized sleep schedule where nodes wake up from their sleep cycle at different times. This is inefficient when broadcast communication is used because the broadcasted frames would need to be stored at the neighbors who are awake and end up being transmitted multiple times. The second drawback of the protocol is the fact that it is vulnerable to the hidden terminal problem where collision can happen at a node if it receives transmissions from two nodes that are not within transmission range of each other (Note that this is not a

TRAMA (Rajendran et al., 2003) is a collision-free TDMA based MAC protocol for sensor networks. TRAMA ensures energy efficiency by avoiding collision during unicast, multicast and broadcast transmissions. In addition, in TRAMA, nodes can switch to a low-power state whenever they are not transmitting or receiving frames to save energy. In TRAMA, a node is elected to transmit within a two-hop neighborhood during each time slot. This mechanism

TRAMA achieves significant energy savings due to: 1. the increased amount of low-power states, 2. the decreased amount of communication since the receiving nodes are indicated a priori, and 3. the significant decrease in collision probability. However, the latency when using TRAMA is longer compared to CSMA as a result of the high percentage of sleep time

Berkeley MAC (B-MAC) (Polastre et al., 2004) is a low frills protocol based on clear channel assessment, it uses low power listening with preamble sampling. The default mode in B-MAC does not include a mechanism to avoid the hidden terminal problem, which could be implemented by higher layers if needed. B-MAC achieves significant energy savings when varying check time, by making the preamble constant and setting the sample rate. However, since the protocol is bare-bone, additional features would have to be implemented at higher

Even though multiple MAC layer protocols provide adequate performance, no single protocol has been chosen as a standard. This is due to the fact that some protocols perform better than others for particular applications, communication pattern, network infrastructures and or network densities. An ideal energy efficient MAC layer protocol for WSNs would use a local schedule for motes to turn to the low-power/off state as a function of their residual energy as well as their sensing schedule. The schedule should aim to maximize the sleep time of the motes while preserving their sensing schedule, local connectivity and balancing their

Frequency detection, generation, modulation and coupling are the responsibility of the physical layer and are explained in detail in the hardware section. Note that when an engineer

layers when needed, which increases the complexity of the system as a whole.

energy levels in order to increase the lifetime of the network as a whole.

problem if WiseMAC is only used as a downlink protocol)

avoids the hidden terminal problem.

(Demirkol et al., 2006).

**4.4.1 Physical layer**

is charged with designing a physical layer, propagation effects due to the ambient conditions must be considered.

### **5. Conclusion and future work**

This chapter reviews the hardware architecture of wireless sensor motes, as well as their protocol stack focusing on power considerations at every level. We conclude that because of the diversity in WSN applications, it is very difficult to derive a universal power efficient architecture both in terms of hardware and software.

As far as the hardware components in WSNs, many advances have been made over the last few years. These improvements include more efficient apertures with better directivity and lower VSWR. The sensor element has been made to become more resolute while power management has improved due to the accessibility of more exotic materials for energy storage. The future holds near perfect antenna with nearly a 1:1 VSWR ensuring most of the energy leaving the system goes where it's designed to propagate. Researchers at Purdue University are working toward ensuring optical sensors are near perfectly efficient with negative refractive metamaterials and photon collection efforts.

In terms of the WSN protocol stack, no one protocol has been adopted as a WSN standard, rather each protocol is designed to efficiently serve one or more WSN applications. The power efficiency of protocols has become the number one constraint in almost every layer of the protocol stack. More work is needed to design and develop protocols that are less application specific and still power efficient.

#### **6. References**


W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, *Energy-Efficient communication*

Power Considerations for Sensor Networks 149

Fei Hu, and Xiaojun Cao, *Wireless Sensor Networks: Principles and Practice*. CRC Press, 2010. C. Intanagonwiwat, R. Govindan and D. Estin, *Directed Diffusion: A scalable and robust*

Sandy Irani, Sandeep Shukla, and Rajesh Gupta, *Power Savings* . Proceedings of ACM Transactions on Algorithms, Vol. 3, No. 4, Article 41, November 2007. Konard Iwanicki, and Maarten van Steen, *On Hierarchical Routing in Wireless Sensor Networks*. Proceedings of IPSN'09, April 13-16, 2009, San Francisco, California, USA. R. E. Kalman, *A new approach to linear filtering and prediction problems*. Transactions. ASME J.

Mauritus Morne, *Reliable Ad-hoc On-demand Distance Vector Routing Protocol*. Proceedings of the

Joanna Kulik, Wendi Rabiner, Hari Balakrishnan, *Adaptive Protocols for Information*

P. Kyasanur, R. R. Choudhury, and I. Gupta. *Smart Gossip: An Adaptive Gossip-based Broadcasting Service for Sensor Networks*. Proceedings of MASS'06, 2006. Andreas Lachenmann, Pedro Jos'e Marr'on, Matthias Gauger, Daniel Minder, Olga Saukh,

Xu, Y., Lee, W. C., Xu, J., and Mitchell, G. (2006). *Processing window queries in wireless sensor*

Christoph Lenzen, Philipp Sommer, and Roger Wattenhofer, *Optimal clock synchronization*

Christoph Lenzen, Philipp Sommer and Roget Wattenhofer, *Optimal Clock Synchronization in Networks*. Proceedings of SensSysÕ09, November 4-9, 2009, Berkeley, CA, USA. P. Levis, A. Tavakoli and S. Dawson-Haggerty, *Overview of Existing Routing Protocols*

Qun Li, Javed Aslam, and Daniela Rus, *Online power-aware routing in wireless ad-hoc networks*.

Liqun Li, Guoliang Xing, Limin Sun, Wei Huangfu, Ruogu Zhou, and Hongsong Zhu,

Proceedings of the ACM MobiSys'11, Washington DC, June 28, 2011. Liu, C., Wu, K., and Pei, J. (2005). *A dynamic clustering and scheduling approach to energy saving*

International Conference on Networking, International Conference on Systems and International Conference on Mobile Communications and Learning Technologies

*Dissemination in Wireless Sensor Networks*. Proceedings of the 5th ACM/IEEE

and Kurt Rothermel, *Removing the Memory Limitations of Sensor Networks with Flash-Based Virtual Memory*. Proceedings of the 2nd ACM SIGOPS/EuroSys European

*networks*. In IEEE International Conference on Data Engineering (ICDEÕ06), Atlanta,

*in network*. Proceedings of the the 7th ACM Conference on Embedded Networked

*for Low Power and Lossy Networks*. IETF ROLL, IETF draft, 14 February 2009,

Proceedings of the 7th Annual International Conference on Mobile Computing and

*Exploiting FM Radio Data System for Adaptive Clock Calibration in Sensor Networks*.

*in data collection from wireless sensor networks*. In Proceedings of the Second Annual

Conference System Sciences, Hawaii, January 2000.

Boston, MA, August 2000.

(ICNICONSMCL'06).

GA, April.

Basic Engin. 82, pages 35-45, 1960.

Mobicom Conference, Seattle, WA, August 1999.

Conference on Computer Systems 2007.

Sensor Systems (SenSys'09).

Networking, 2001.

Draft-ietf-roll-protocols-survey-07.

*protocol for wireless sensor networks*. Proceedings of the Hawaii International

*communication pradigm for sensor networks*. Proceedings of the 6th annual ACM/IEEE International Conference on Mobile Computing and Networking (MobiComÕ00),

Crossbow Technologies, *http://www.xbow.com/*


20 Will-be-set-by-IN-TECH

Ilker Demirkol, Cem Ersoy, and Fatih Alagoz, *MAC Protocol for Wireless Sensor Networks: a*

A. P. Dempster, *A generalization of Bayesian inference*. J. Royal Stat. Soc., Series B 30, pages

Isabel Dietrich and Falko Dressler, *On the lifetime of wireless sensor networks*. ACM Transactions

Draft-IETF-MANET-DYMO-mib-04: http://tools.ietf.org/html/draft-ietf-manet-dymo-mib-04

E. H. Elhafsi, N. MItton, D. Simplot-Ryl, *End-to-End Energy Efficient Geographic Path Discovery*

Jeremy Elson, Lewis Girod, and Deborah Estrin, *Fine-grained network time synchronization using reference broadcasts*. Proceedings of the ACM OSDI'02, Boston, MA, December 2002. Mehdi Esnaashari and M.R. Meybodi. *Data Aggregation in Sensor Networks using Learning*

S. Ganeriwal, R. Kumar, and M Srivastava, *Timing Sync Protocol for Sensor Networks*.

J.V. Greunen, and J. Rabaey, *Lightweight Time Synchronization for Sensor Networks*. Proceedings

S. Grime, H.F. Durrant-Whyte, *Data fusion in decentralized sensor networks*. Control Eng.

N. Guilar, A. Chen, T. Kleeburg, and R. Amirtharajah, *Integrated Solar Energy Harvesting and*

I. Gupta, D. Riordan, and S. Sampalli, *Cluster-head election using fuzzy logic for wireless sensor*

Heinzelman, W., Chandrakasan, A., and Balakrishnan, H. (2000). *Energy-efficient*

W.B. Heinzelman, A. P. Chandrakasan, and H. Blakrishnan, *An Application-Specific*

W. Heinzelman, J. Kulik, and H. Balakrishnan, *Adaptive protocols for information dissemination*

Research Conference (CNSR'05). IEEE, Halifax Canada, pages 255-260. Tian He, John Stankovic, Chenyang Lu, and Tarek Abdelzaher, *SPEED: A Real-Time Routing*

International Conference on System Science (HICSS Ô00), January.

Communications, vol. 1, no. 4, pp. 660-670, October 2002.

of the 2nd ACM International Conference on Wireless Sensor Networks and

*Storage*. Proceedings of the International Symposium of Low Power Electronics and

*networks*. Proceedings of the 3rd Annual Communcation Networks and Services

*communication protocol for wireless microsensor networks*. In Proceedings of 33rd Hawaii

*Protocol Architecture for Wireless Microsensor Neworks*. IEEE Transactions on Wireless

*in wireless sensor networks*. Proceedings of the 5th annual ACM/IEEE International Conference on Mobile Computing and Networking (MobiComÕ99), Seattle, WA,

*Infrastructure Wireless Sensor Networks*. Proceedings of the Ninth IEEE Symposium on Computers and Communication, ISCCÕ04, pages 244-251, Alexandria, Egypt, June

*With Guaranteed Delivery in Ad Hoc and Sensor Networks*. 19th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC).

Crossbow Technologies, *http://www.xbow.com/*

205-247, 1968.

2004.

*Survey*. IEEE Communications Magazine, 2006.

on Sensor Networks, Vol. 5, No. 1, Article 5, February 2009.

Cannes, France: IEEE, 15-18 Septembr 2008, pp. 1-5.

Applications (WSNA'03), San Diego, CA, September 2003.

*Automata*. Wireless Networks 20

Practices, 2(5):849-63, 1994.

*Protocol for Sensor Networks*.

August 1999.

Design (ISLPED'06), October 2006.

Proceedings of the ACM SenSys'03, 2003.

Draft-IETF-MANET-NHDP-15: http://tools.ietf.org/html/draft-ietf-manet-nhdp-15 Draft-IETF-MANET-olsrv2-12: https://datatracker.ietf.org/doc/draft-ietf-manet-olsrv2/ A. El-Hoiydi and J.-D. Decotignie, *WiseMAC: An Ultra Low Power Protocol for the Downlink of*


RFC 3561: http://www.ietf.org/rfc/rfc3561.txt RFC 3626:http://www.ietf.org/rfc/rfc3626.txt RFC 3684: http://www.ietf.org/rfc/rfc3684.txt RFC 4728: http://www.ietf.org/rfc/rfc4728.txt RFC 4861: http://tools.ietf.org/html/rfc4861

Beach, CA October 2001.

Sensor Networks, 2009.

5, pages:1122-1135, March 2007.

Systems, November 2003.

and Tutorials.

5-26

Information Force, McLean, VA 2001.

Kay Romer, *Time synchronization in ad hoc networks*. Proceedings of the ACM MobiHoc'01, Long

Power Considerations for Sensor Networks 151

Rosemark, R., and Lee, W. C. (2005). *Decentralizing query processing in sensor networks*. In The

C. Schurgers and M.B. Srivastava, *Energy efficient routing in wireless sensor networks*. MILCOM

G. Shafer, *A Mathematical Theory of Evidence*. Princelton University Press, Princeton, NJ 1976. Vladimir M. Shalaev, Wenshan Cai, Uday K. Chettiar, Hsiao-Kuan Yuan, Andrey K. Sarychev,

Philipp Sommer and, Roger Wattenhofer, *Gradient Clock Synchronization in Wireless Sensor*

Soro, S., and Heinzelman, W. (2005). *Prolonging the lifetime of wireless sensor networks via*

A. Tsymbal, S. Puuronen, and D. W. Patterson, *Ensemble feature selection with the simple Bayesian*

T. van Dam, and K. Langendoen, *An adaptive energy-efficient mac protocol for wireless sensor*

Natalia Vassileva, Francisco Barcelo-Arroyo, *A Survey of Routing Protocols for Maximizing the*

Virrankoski, R., and Savvides, A. (2005). *TASC: Topology adaptive spatial clustering for sensor*

Thomas Watteyne, Maria Grazia Richichi, *From MANET to IETF ROLL Standardization: A*

Jon Wilson, *Sensor Technology Handbook*. Elsevier, ISBN: 0-7506-7729-5, December 2004. Winter, J., Xu, Y., and Lee, W. C. (2005). *Energy efficient processing of K nearest neighbor queries in*

*classification*. Information Fusion 4, 2, June, pages 87-100.

and its Applications, Vol. 2, No. 3, July 2008.

systems, Washington, DC, November.

and Services (MobiquitousÕ05), San Diego, CA, July (pp. 270Ð280).

*metamaterials*. Optics Letters, Vol. 30, Issue 24, pages 3356-3358, 2005. S. Singh and C.S. Raghavendra, *Power aware multi-access protocol with signaling for ad hoc*

Second International Conference on Mobile and Ubiquitous Systems: Networking

Proceedings on Communications for Network-Centric Operations: Creating the

Vladimir P. Drachev, and Alexander V. Kildishev, *Negative index of refraction in optical*

*networks*. ACM Computer Communication Review Vol. 28 No. 3 (July 1998) pages.

*Networks*. Proceedings of the International Conference on Information Processing in

*unequal clustering*. In Proceedings of the 5th International Workshop on Algorithms for Wireless, Mobile, Ad Hoc and Sensor Networks (IEEE WMAN Ô05), April. K. Stewart, and S. Tragoudas, *Managing the power resources of sensor networks with*

*performance considerations*. Computer Communications Journal, Volume 30, Number

*networks*. Proceedings of the First ACM Conference on Embedded Networked Sensor

*Lifetime of Ad Hoc Wireless Networks*. International Journal of Software Engineering

*networks*. In Second IEEE International Conference on Mobile Ad Hoc and Sensor

*Paradigm Shift in WSN Routing Protocols*, submitted to IEEE Communications Surveys

*location-aware sensor networks*. In The Second International Conference on Mobile and

IEEE Communications Society Conference on Sensor and Ad Hoc Communications and Networks (SECONÕ05), Santa Clara, California, USA, September.


22 Will-be-set-by-IN-TECH

Lotfinezhad, M., and Liang, B. (2004). *Effect of partially correlated data on clustering in wireless*

Saoucene Mahfoudh, and Pascale Minet. *Energy-aware Routing in Wireless Ad Hoc and Sensor*

Alan Mainwaring, Joseph Polastre, Robert Szewczyk, David Culler, and John Anderson.

M. Marina, and S. Das, *On-demand Multipath Distance Vector Routing in Ad Hoc Networks*.

M. Maroti, B. Kusy, G. Simon, and A. Ledeczi, *The Flooding Time Synchronization Protocol*.

Morteza Maleki, Karthik Dantu, and Massoud Pedram *Power-aware Source Routing Protocol for*

Eduardo Nakamura, and Alejandro Frery. *Information Fusion for Wireless Sensor Networks:*

Evgenii E. Narimanov, and Alexamder V. Kildishev, *Optical black hole: broadband omnidirectional*

D. Petriovic, R. C. Shah, L. Ramchandran, and J. Rabaey, *Data funneling: Routing with*

Kris Pister, *Autonomous sensing and communication in a cubic millimeter*.

Joseph Polastre, Jason Hill, and David Culler, *Versatile low power media access for wireless sensor*

Octavian Postolache, Pedro Silva Girao, and Jose Miguel Dias Pereira, *Non-Volatile Memory*

V. Rajendran, K. Obraczka, J. J. Garcia-Luna-Aceves, *Energy-Efficient, Collision-Free Medium*

Http://www-bsac.eecs.berkeley.edu/ pister/SmartDust/

Computing Conference (IWCMCÕ10). June 2010.

pages 14-23, IEEE Computer Society Press, 2001.

*light absorber*. Appl. Phys. Lett. 95, 041106 (2009).

IEEE, Anchrage, AK, pages 156-162.

2003, Los Angeles, California, USA.

RFC 1142: http://tools.ietf.org/html/rfc1142 RFC 2328:http://www.ietf.org/rfc/rfc2328.txt RFC 2453: http://tools.ietf.org/html/rfc2453

publishers, April 2010.

*networks*. Proceedings of SenSys'04, 2004.

(SenSys'04), Baltimore, Maryland, 2004, pages: 39-49.

and Networks (SECONÕ05), Santa Clara, California, USA, September. V. Loscri, G. Morabito, and S. Marano, *A Two-Level Hierarchy for Low-Energy Adaptive Clustering*

25-28, 2005.

28, 2002, Atlanta, Georgia.

California, USA.

3, 2007.

NTP: http://www.ntp.org/

OLSR: RFC 3626

IEEE Communications Society Conference on Sensor and Ad Hoc Communications

*Hierarchy*. Proceedings of the Vehicular Technology Conference (VTC'05), September

*sensor networks*. In Proceedings of the IEEE International Conference on Sensor and Ad hoc Communications and Networks (SECON), Santa Clara, California, October.

*Networks*, Proceedings of the 6th International Wireless Communications and Mobile

*Wireless Sensor Networks for Habitat Monitoring*. Proceedings of WSNA'02, September

Proceedings of the 2001 IEEE International Conference on Network Protocols (ICNP),

Proceedings of the 2nd ACN Conference on Embedded Networked Sensor Systems

*Mobile Ad Hoc Networks*. Proceedings of the ISLPED'02, August 12-14, 2002, Monterey,

*Methods, Models, and Classifications*, Computing Surveys (CSUR), Volume 39, Issue

*aggregation and compression for wireless sensor networks*. Proceedings of the first IEEE International Workshop on Sensor Network Protocols and Applications (SNPA'03).

*Interface Protocols for Smart Sensor Networks and Mobile Devices*. Data Storage, InTech

*Access Control for Wireless Sensor Networks*. Proceedings of SenSysÕ03, November 5-7,

RFC 3561: http://www.ietf.org/rfc/rfc3561.txt

RFC 3626:http://www.ietf.org/rfc/rfc3626.txt

RFC 3684: http://www.ietf.org/rfc/rfc3684.txt

RFC 4728: http://www.ietf.org/rfc/rfc4728.txt


**0**

**7**

**Review of Optimization Problems**

Ada Gogu1, Dritan Nace1, Arta Dilo2 and Nirvana Meratnia2

Wireless Sensor Networks (WSNs) are an interesting field of research because of their numerous applications and the possibility of integrating them into more complex network systems. The difficulties encountered in WSN design usually relate either to their stringent constraints, which include energy, bandwidth, memory and computational capabilities, or to the requirements of the particular application. As WSN design problems become more and more challenging, advances in the areas of Operations Research (OR) and Optimization are

This study is concerned with topics relating to network design (including coverage, topology and power control, the medium access mechanism and the duty cycle) and to routing in WSN. The optimization problems encountered in these areas are affected simultaneously by different parameters pertaining to the physical, Medium Access Control (MAC), routing and application layers of the protocol stack. The goal of this study is to identify a number of different network problems, and for each of these network problems to examine the underlying optimization problem. In each case we begin by presenting the basic version of the network problem and extend it by introducing new constraints. These constraints result mainly from technological advances and from additional requirements present in WSN applications. For all the network problems discussed here a wide range of algorithms and protocols are to be found in the literature. We cite only some of these, since we are concerned more with the network optimization problem itself, together with its different versions, than with a state of art of methods for solving it. Moreover, the cited methods have originated in a variety of disciplines, with approaches ranging from the deterministic to the opportunistic, including computational geometry, linear, nonlinear and dynamic programming, metaheuristics and heuristics, game theory, and so on. We go on to discuss the complexity inherent in different optimization problems, in order to give some hints to WSN designers facing new but similar scenarios. We try to highlight distributed solutions and information that is required to implement these schemes. For each topic the general

**1. Introduction**

becoming increasingly useful in addressing them.

presentation scheme is as follows: i) Present the network problem

ii) Identify the relevant optimization problem

**in Wireless Sensor Networks**

<sup>1</sup>*Université de Technologie de Compiègne*

<sup>2</sup>*University of Twente*

<sup>2</sup>*The Netherlands*

<sup>1</sup>*France*

Ubiquitous Systems: Networking and Services (MobiquitousÕ05), San Diego, CA, July (pp. 281Ð292).

