Investigating the Energy Sink-Hole Problem in Connected k-covered Wireless Sensor Networks

Transcription

1 Investigating the Energy Sink-Hole Problem in Connected k-covered Wireless Sensor Networks Habib M. Ammari Wireless Sensor and Mobile Ad-hoc Networks (WiSeMAN) Research Lab Department of Computer and Information Science, University of Michigan-Dearborn, Dearborn, Michigan, USA Abstract: In immobile wireless sensor networks with constant data reporting, the sensors nearer the sink are responsible for forwarding data to it on behalf of all other sensors in the network. Those sensors suffer from a severe batter power depletion problem, also known as the energy sink-hole problem. In this paper, we study the above problem in duty-cycled connected k- covered wireless sensor networks, where each point in a field of interest is covered by at least k sensor. In order to change the neighbors of a sink over time, our solutions suggest the use of mobile proxy sinks that collect data from source sensors and drop them off at an immobile sink. Our proposed three-tier architecture has immobile source sensors, immobile sinks, and mobile proxy sinks. First, we present our fundamental results for the design of duty-cycled connected k- covered wireless sensor networks. Second, we provide the first formal analysis of the performance of joint mobility and routing in this type of network. Precisely, we investigate the best mobility strategy of mobile proxy sinks to minimize the total energy consumption for data collection. Third, we propose joint mobility and routing schemes based on the number of immobile sinks and mobile proxy sinks. We provide a thorough analytical model for our schemes. Finally, we evaluate their performance by simulation. Our results show their significant improvement of the network lifetime compared to a solution without mobile proxy sinks. Keywords: Wireless sensor networks, k-coverage, data collection, energy sink-hole problem, mobility. 1. INTRODUCTION A wireless sensor network (WSN) consists of tiny, low-powered sensors that communicate wirelessly with each other and report their data to the sink (i.e., central gathering point), for further analysis and processing. There are two fundamental concepts in the design of WSNs, namely coverage and connectivity. In fact, data collection and routing suggest that a deployment field be covered and that the network be connected [31]. However, some applications require k-coverage [13], [19], [22], where each point in a field is covered by at least k sensors while all those sensors are connected. This problem is referred to as connected k-coverage in WSNs. The challenge is to select a minimum subset of sensors to remain active in order to k-cover a field and ensure connectivity between all active sensors. In this type of duty-cycled WSN, connectivity is time-varying due to the fact that sensors may be on or off. Also, some sensors may deplete their energy and die. Thus, routing on duty-cycled sensors is challenging as those ones that are selected as next forwarders may not be on or have depleted all of their energy when data reached them. Sensors power depletion may have a serious problem that affects the network performance. In particular, this problem gets aggravated in the case of immobile WSNs when some specific sensors deplete their energy. Indeed, the sensors nearer the sink are very critical to this problem as they act as the points of contact between the sink and the rest of the sensors in the network. It is easy to check that the sensors around the sink severely suffer from a battery power depletion problem. Those sensors are responsible for forwarding data on behalf of all other sensors in the network to the immobile sink. Thus, the death of those sensors may yield a coverage hole around the immobile sink, which prohibits the data from reaching it. This phenomenon is known as the energy sink-hole problem. 1.1 Impact of the Energy Sink-Hole Problem It is well known that homogeneous WSNs using only immobile source sensors and immobile sinks suffer from the energy sink-hole problem [15], [20]. Given that k-covered WSNs are dense in nature, the energy sink-hole problem may have negative impact on those sensing applications that require k-coverage of their sensor deployment fields. Precisely, those applications will suffer from some or all of the following problems: The area surrounding the immobile sink may not be k-covered as those sensors near the sensors may have depleted their energy. Therefore, the level of k-coverage required by the sensing application may not be achieved in the whole field. The entire network may be split into at least two disconnected sub-nets. In particular, the immobile sink may be disconnected from the rest of the network. It is worth noting that connectivity to the sink is the true metric that should be considered when considering network connectivity in WSNs. In fact, if the sink is not reachable by the sensors, it is as if the network does not have any existence since no sensed data can reach the immobile sink. 1

2 1.2 Motivations and Problem Formulation In this paper, we attempt to solve the energy sink-hole problem in connected k-covered WSNs. Our study of this problem is motivated by the existence of a wide range of real-world sensing applications that deploy immobile sensors and sinks. It is necessary that this problem be addressed and energy-efficient solution be provided for the network operational effectiveness and lifetime elongation. We believe that the latter depend on two design decisions, namely whether the sensors are homogeneous or heterogeneous, and whether they are mobile or immobile. To solve the energy sink-hole problem, we deploy heterogeneous sensors that have different initial energy. This helps extend the network reliability and lifetime [18], [27]. Also, we use mobile sinks in order to achieve balanced load among all the sensors in the network. We expect that the joint use of sensor heterogeneity and sink mobility be able to efficiently address the problem of energy sink-hole in duty-cycled connected k-covered WSNs. In this paper, we propose to address the following three questions: (1) What would be the sensor deployment architecture and the nature of all participating components so as to avoid the energy sink-hole problem in dutycycled connected k-covered WSNs? (2) What would be the mobility trajectory of the sink to ensure uniform energy consumption of all the sensors and cope with the problem of energy sinkhole in duty-cycled connected k-covered WSNs? (3) How would data be collected from source sensors by mobile sinks and dropped off at the immobile sinks to minimize energy consumption? 1.3 Major Contributions Our contributions in this paper can be summarized as follows. We propose a three-tier architecture that forms the basis for addressing the energy sink-hole problem in duty-cycled connected k-covered WSNs. Precisely, it uses heterogeneous sensors that differ by their initial energy. The first layer has a set of immobile source sensors (i.e., data generating points) that have the same capabilities and are densely and randomly deployed to achieve k-coverage of a circular field. The second layer has one or more immobile sinks (i.e., central gathering points) that have infinite sources of energy. These sinks are positioned at specific locations in the circular field. The third layer contains one or multiple mobile sinks, called mobile proxy sinks. These sinks have the same capabilities, but are more powerful than the source sensors in terms of their initial energy. Their sole task is to collect data from source sensors and drop them off at the immobile sinks. We use our architecture for data collection using joint mobility and routing on top of energy-efficient connected k-coverage configurations. We propose four data collection schemes depending on the numbers of immobile sinks and mobile proxy sinks. First, we divide a circular field into concentric circular bands. Then, we determine the best mobility strategy of a mobile proxy sink so as to minimize the average total energy consumption of all the source sensors. Second, for each data collection scheme, we give the placement of immobile sinks and mobile proxy sinks, and show how the latter collect data from source sensors and deliver them to the former. Third, we provide a rigorous analysis of the performance of each scheme and corroborate it with simulation results. We show that our proposed schemes significantly improve the network lifetime compared to a solution without mobile proxy sinks. To the best of our knowledge, this is the first work that studies the energy sink-hole problem in duty-cycled connected k-covered WSNs and provides a thorough analysis of joint mobility and routing schemes to cope with it. The rest of this paper is organized as follows. Section 2 presents our network and energy models. Section 3 gives an overview of related work while Section 4 summarizes our fundamental results regarding the connected k-coverage problem in WSNs. Section 5 discusses the energy sink-hole problem, while Section 6 describes our three-tier architecture and discusses our proposed solutions to this problem in duty-cycled connected k-covered WSNs. Section 7 evaluates the performance of our joint mobility and routing schemes. Section 8 concludes the paper. 2. DEFINITIONS AND MODELS In this section, we give some definitions and specify our network and energy models to address the problem of energy sink-hole in connected k-covered WSNs. 2.1 Terminology Definition 1: A WSN is said to be homogeneous if it is composed of sensors that have the same sensing, processing, communication, and energy capabilities. Otherwise, it is said to be heterogeneous. Definition 2: A communication neighbor set of a sensor s i is the set of sensors located in the communication range of s i. Definition 3: The sensing neighbor set of a sensor s i consists of all sensors in the sensing range of s i. Definition 4: A point in a field of interest is k-covered if it is covered (sensed) by at least k sensors. A field of interest is k-covered if each point in it is k-covered. Definition 5: The width of closed convex area is the maximum of the distances between any pair of parallel lines that bound it. Definition 6: A mobile proxy sink is a mobile sensor that only collects data from source sensors and delivers it to an immobile sink when it comes close by. 2

3 2.2 Network Model We consider static WSNs with constant data reporting to the sink, where all the source sensors and sink are static. Moreover, all those source sensors are randomly and uniformly distributed in a circular field of diameter D. In addition, we assume that all the sensors and the sink are aware of their geographic locations via some localization technique [4]. Furthermore, all the source sensors are supposed to be homogeneous, i.e., have the same sensing range, communication range, and initial energy. Also, the sensing and communication ranges of a sensor s i are represented by disks of radii r and R, respectively. Besides, we consider a deterministic sensing model, where a point p in a field is covered by a sensor s i if,, where, stands for the Euclidean distance between p and s i. 2.3 Energy Model We assume that the energy consumption of the sensors is due to data reception, transmission, and mobility. According to [10], the energy spent in transmitting one message of size bits over a distance d, is given by = + = + where =50 10 is the electronics energy, is the transmitter amplifier =10 for =2, and =13 10 for 3, and is the path-loss exponent 2 4 [10]. Also, the energy spent in receiving one message is given by: = Hence, the total energy consumed when a sensor s i transmits a message over a distance d to s j, and s i receives it, is given by = +2 Following [23], the energy spent due to mobility, denoted by, is computed as = where is the energy cost for a mobile sensor to move one unit distance, and is the total distance traveled by the mobile sensor. We assume = RELATED WORK In this section, we review a sample approaches for connected coverage in WSNs. We also discuss existing work on the energy sink-hole problem in WSNs. 3.1 Connected k-coverage The problem of joint coverage and connectivity in WSNs was originally introduced by Xing et al. [25]. They proposed the first joint coverage and connectivity protocol in WSNs, called connected coverage protocol (CCP), to provide different degrees of full coverage of a convex region [25]. Yang et al. [26] formalized the k-connected coverage set problem in terms of linear programming and proposed a non-global solution to it. In [11], Huang et al. studied the relationship between coverage and connectivity of sensor nets and proposed distributed protocols to ensure both of them. Bai et al. [2] proposed an optimal deployment strategy to achieve coverage and 2-connectivity regardless of the relationship between sensing and communication radii of the sensors. Gupta et al. [9] proposed centralized and distributed algorithms for connected sensor cover so the WSN self-organizes its topology in response to a query and activate the necessary sensors to process the query. Zhang and Hou [29] proposed an optimal geographical density control protocol to keep a small number of sensors active regardless of the ratio of the sensors communication range to their sensing range. 3.2 Energy Sink-Hole Problem The energy sink-hole problem in WSNs has gained relatively less attention in the literature. It is worth noting that this problem was originally addressed by Guo et al. [8]. They proposed an energy-balanced transmission scheme that adjusts the ratio between direct transmission and next-hop transmission. Precisely, sensors far away from the sink send larger percent of data to the next hop, while sensors near the sink send more data directly to the sink. Zhang et al. [30] also exploited this combination of hop-by-hop transmission and direct transmission to find a trade-off between them. Efthymiou et al. [5] proposed a probabilistic data propagation algorithm for balancing energy consumption among all sensors. Powell et al. [21] used the probabilistic data propagation algorithm in [5] and proved that there is a relationship between energy balancing and lifespan maximization [12]. Leone et al. [14] considered non-uniform sensor distribution and proposed a blind algorithm that computes a solution to the energy-balancing problem on-line without prior knowledge on the occurrences of the events. Li and Mohapatra [15] proposed an analytical model and investigated the effectiveness of a few approaches for mitigating the energy hole problem. Olariu and Stojmenovic [20] proved that energy-efficient routing can be guaranteed when the coronas of a circular field have the same width, but this leads to uneven energy depletion of sensors. To cope with this, they computed the widths of coronas and their number. They also proved that uneven energy depletion is unavoidable for the free space model but can be prevented for the multi-path model. Lian et al. [16] showed that up to 90% of the total initial energy is unused due to the static WSN model with uniformly distributed homogenous sensors and a stationary sink. Wu et al. considered a non-uniform node distribution to achieve balanced energy depletion [24]. Luo and Hubaux [17] proposed a data collection protocol for WSNs that makes use of multi-hop routing and base station mobility to solve the energy-sink-hole problem. 3

4 4. CONNECTED k-coverage In our study of the energy sink-hole problem in dutycycled connected k-covered WSNs, we exploit previous results on connected k-coverage in WSNs and a randomized distributed protocol [1], which are based on Helly s Theorem [3] a fundamental result of the theory of convexity. In this section, we give a brief summary of these results and distributed protocol. 4.1 Fundamental Results Helly s Theorem [3]: Let E be a family of convex sets in R n such that for m n+1 any m members of E have a non-empty intersection. Then, the intersection of all members of E is non-empty. Lemma 1 [1] is an instance of Helly s Theorem [3] in a two-dimensional space that characterizes the intersection of k sensing disks, with n = 2 and k = m. Lemma 1 [1]: Let k 3. The intersection of k sensing disks is not empty if and only if the intersection of any three of those k sensing disks is not empty. Based on Lemma 1, Lemma 2 [1] gives a sufficient condition for k-coverage of a field. Lemma 2 [1]: Let k 3. A field is k-covered if any Reuleaux triangle of width r (or slice as shown in Figure 1) in the field contains at least k active sensors, where r is the radius of the sensors sensing disks. Theorem 1 [1]: Let k 3. The sensor spatial density that guarantees k-coverage of a field is given by 6,= with r being the radius of the sensors sensing disks. 4.2 Distributed k-coverage Protocol Each sensor s i runs a k-coverage checking algorithm [1] to check whether its sensing disk is k-covered. Precisely, s i divides its sensing disk into six overlapping slices as shown in Figure 3 such that two adjacent slices intersect in a lens. For each slice, s i computes its coverage degree. To this end, s i checks whether the number of active sensors in each lens of two adjacent slices, including itself, is equal to k. Based on this result, s i activates a necessary number of its sensing neighbors to k-cover its sensing disk. For more details about this distributed randomized connected k-coverage protocols, the interested reader is referred to [1]. Figure 5 Figure 6: Plot of Figure 1 [1] Figure 2 [1] Figure 3 [1] Figure 4 [1] Figure 2 shows a slicing grid while Figure 3 shows that the sensing disk of a sensor can be covered by six overlapping slices of width r. Thus, the minimum overlap area of two adjacent slices forms a lens as shown in Figure 4. Lemma 3 characterizes k-coverage. Lemma 3 [1]: k active sensors located in the lens of two adjacent slices can k-cover both slices. Theorem 1 [1] computes the sensor spatial density needed for k-coverage of a field based on Lemma ENERGY SINK HOLE PROBLEM In this section, we study the basic architecture that has only one immobile sink, zero mobile proxy sink, and densely deployed source sensors to k-cover the field. As stated earlier, this architecture suffers from the energy sink-hole problem. We assume that the source sensors report their data constantly to the sink using a short-path routing protocol [7]. We will show that the source sensors around the sink consume higher energy compared to all other source sensors in the network. In order to compute the maximum average energy consumption of the source sensors, we use a model that is similar to the one in [6]. Precisely, the average energy consumption of a node located in an area of size that forwards traffic for other nodes located in another area of size is proportional to + /. We focus on the source sensors within a distance from the sink, where stands for the radius of the communication range of the source sensors and is an infinitesimal value, as shown in Figure 5. More specifically, we consider a circle of radius around the immobile sink. It is clear that includes the source sensors that are actively forwarders data on behalf of all other source sensors to. The 4

5 number of source sensors inside is, while that of source sensors outside is, where is the radius of the circular deployment field and is the sensor spatial density to k-cover the field. Let be the radius of an infinitesimal circular region A whose area is =, where 0 and 0 2. The average distance between a source sensor in and the immobile sink is computed as follows: = 1 = 1 = 2 3 The energy consumption per source sensor in is given by =, +, where, is the average energy consumption per source sensor in C σ to directly send its data to the immobile sink, and, is the average energy consumption per source sensor in to forward a subset of data packets originated from source sensors in = to the sink. Using the energy model in Section 2, we obtain, = 1 = , = Thus, the energy consumption per source sensor using the base protocol is given by = Figure 6 shows that increases significantly as we approach the center of the sensor field (i.e., location of the sink). It is clear that all sensed data will be forwarded by sensors whose distance from the sink is at most equal to R = 300m. 6. JOINT MOBILITY AND ROUTING In this section, we propose four different approaches for data collection to address the problem of energy sink-hole problem in connected k-covered WSNs. We assume that there is at least one immobile sink, where the collected data is stored. Also, we suppose that any immobile sink has an infinite source of energy [27]. Specifically, these approaches are based on the number of immobile sinks and that of mobile proxy sinks. First, we give a decomposition of a deployment field. Then, we investigate the optimum mobility of a mobile proxy sink to minimize the average total energy consumption of the source sensors. Finally, we discuss each of our data collection approaches in details and compute their average total energy consumption. 6.1 Sensor Field Deployment Decomposition We divide a circular field of interest of diameter D into n concentric circular bands of width R as shown in Figure 7, where n = D/R with R being the radius of the sensors communication range. Our proposed data collection schemes are based on this decomposition. Particularly, the placement of the immobile sinks and the mobility of mobile proxy as well as their numbers sinks depend on this deployment field decomposition. Figure 7 Figure Optimum Proxy Sink Mobility Trajectory To ensure energy-efficient data collection, we need to specify the mobility trajectory of the mobile proxy sink and how data is being collected from source sensors. The mobile proxy sink will have circular trajectories interrupted by short linear moves. First, we determine the optimum mobility trajectory of the mobile proxy sink that yields the minimum energy consumption when collecting data. Theorem 2 states this result under the energy model specified in Section 2.3. Theorem 2: Let = 1. The optimum proxy sink mobility trajectory inside a band b i of width that yields the minimum energy consumption during data collection, corresponds to a circle of radius given by = Proof: Consider Figure 8. Assume that a mobile proxy sink mps 1 moves inside band b i on the perimeter of a circle of radius + denoted by,+, where 0. We distinguish two groups of source sensors in band b i. The first group is in band b i,1 of interior and exterior radii and +, respectively. The second group belongs to band b i,2 of interior and exterior radii + and +, respectively. That is, band b i is split into bands b i,1 and b i,2. Let N 1 and N 2 be the numbers of source sensors in bands b i,1 and b i,2, respectively. Based on spatial sensor density stated in Theorem 1 above to ensure k-coverage of a field,, and, can be computed as follows:, =, =+, =, =+ + where, and, stand for the areas of bands b i,1 and b i,2, respectively, and = 1. 5

6 To minimize the average energy consumption due to data transmission and reception, it is necessary that both subsets of sensors in bands b i,1 and b i,2 consume the same amount of energy. Given that the energy consumption depends on the number of source sensors transmitting their data, a balanced energy consumption between bands b i,1 and b i,2 is achieved when, =,. Hence, we have the equality: + =+ + Thus, the optimum value of, denoted by, which yields the minimum total energy consumption by all source sensors in both bands b i,1 and b i,2, is given by = Thus, the radius =+ of the mobility circle, of mobile proxy sink mps 1 is given by =+ = Table 1: Values of depend on Figure 9: vs. b i Notice that depends on the band being visited by mps 1. As increases, b i,1 and b i,2 tend to have the same area, and, thus, the same number of source sensors. tends to decrease and reach a value close to. Table 1 gives the values of for a few bands, while Figure 9 plots as a function of band b i Immobile Sink 1 Mobile Proxy Sink There are various scenarios for a mobile proxy sink to collect data. Next, we discuss two specific ones. Figure 10 Figure 11 Scenario 1 (Single band-based data collection): In this scenario, the mobile proxy sink mps 1 moves inside each band to gather data from the sensors located inside that band. More precisely, mps 1 starts its movement at one location l s in band b i, moves on the perimeter of a circle of radius + inside b i, and returns to l s. Once mps 1 revisits l s, it goes to the next band by moving through a linear trajectory (see Figure 10). We assume that mps 1 moves at a constant speed between 0 and v max, and stops at some locations to collect data. The mobile proxy sink mps 1 follows the same trajectory pattern until it visits all bands b 1,, b n and collects data from their respective source sensors. Let us compute the average transmission distance used by all the source sensors. As shown in Figure 11, each source sensor s i sends its data to the mobile proxy sink mps 1 only when the segment connecting the respective locations of s i and mps 1, say and, is orthogonal to the tangential passing by located at the perimeter of the mobility circle,+. In fact, this scenario corresponds to the minimum transmission distance s i can use to minimize its energy consumption when sending its data. While source sensors in b i,1 transmit their data to mps 1 over distances ranging between 0 and, those in b i,2 send their data to mps 1 over distances ranging between 0 and. Lemma 2 computes the average distance between source sensors in band b i,1 and mps 1. Lemma 2: The average Euclidean distance,,,, between a source sensor in band b i,1 whose width is, and the mobile proxy sink mps 1 is computed as,, = 1 3 Figure 12 6

7 Proof: Consider band b i,1 whose width is equal to. If we unfold b i,1, we get a diagram that can be modeled by the segment 0,, where each source sensor is located at position 0, whereas mps 1 is positioned at location =0, with, as shown in Figure 12. All the locations of the source sensors form a uniform distribution. We pick a source sensor s i randomly located at =0,. The average Euclidean distance between and is given by,,,, =,,,, = = 1 2 Given that =0,, =0,, and, we deduce that =. Thus, we have,, = = 1 6 We conclude that,,, =. Corollary 1 states the average distance between a source sensor s i in band b i,2 and mps 1, using Lemma 1. Corollary 1: The average Euclidean distance,,,, between a source sensor in band b i,2 whose width is, and the mobile proxy sink mps 1 is given by,, = Let,, be the average energy consumption that is due to the transmission of data by all the source sensors located inside band b i,1 and its reception by the mobile proxy sink mps 1. Knowing the number of source sensors in band b i,1 and the average distance,,, the energy,, is given by,, =,, = +,, +2 = Likewise, we denote by,, the average energy consumption, which results from the transmission of data by all the source sensors located inside band b i,2 and its reception by the mobile sink mps 1. Based on the number of source sensors in band b i,2 and the average distance,,, the energy,, is computed as follows:,, =,, = + +,, +2 = Thus, the total average energy consumption,,, due to data transmission by all source sensors located inside band b i and its reception by mps 1, is given by:, =,, +,, Now, the energy spent by mps 1 when moving on the circle,+ located inside band b i is given by,, = 2+ Also, the energy spent by mps 1 to move up to the next band b i+1 using a rectilinear trajectory of length R is,, = Moreover, after collecting all data from all sources sensors in all bands, mps 1 should deliver it to the static sink located in band b1 at the center of the field. Let,, be the energy consumed by mps 1 to move from band b n to band b 1, where it drops off all collected data to the sink. As It is clear that mps 1 has to traverse n 1 bands (b n-1,,b 1 ), each of which has width equal to R. The distance that mps 1 has to travel while in band b n is equal to. Hence, we have,, = 1+ The total energy consumption is due to the mobility of mps 1 and data transmission by source sensors in band b i to mps 1, and its reception by mps 1. Thus, the total average energy consumption,, per round of data collection from all n bands b 1,,b n, is given by =,, + +, +,,,, Scenario 2 (Adjacent bands-based data collection): We assume that mps 1 moves on the perimeters of the bands. During its mobility on the perimeter of band b i, mps 1 collects data from the source sensors located in the two adjacent bands b i and b i+1. Then, it takes a linear trajectory to move up to band b i+2 in order to collect data from source sensors in bands b i+2 and b i+3. As can be seen, in every round of movement for data collection, mps 1 gathers data from a new pair of adjacent bands. The mobile proxy sink mps 1 repeats the same movement strategy until it collect data from source sensors in all bands b 1,,b n. Corollary 2: The average Euclidean distance,, between a source sensor located in band b i whose width is and the mobile proxy sink mps 1 moving on the outer circle of band b i is computed as follows: = 7

8 Let us compute the total energy consumption due to data collection from the source sensors in all bands b 1,,b n as well as the proxy sink mobility. We distinguish two cases depending on whether the number of bands n is odd or even. Case 1: n is even (n = 2m) The mobile proxy sink mps 1 collects data from a pair of adjacent bands. Let, be the average energy consumption for data collection from all the source sensors located in band b i. We have:, = Thus, the total average energy consumption,, which is due to the transmission of data by the source sensors located in all bands b 1,, b n and its reception by mps 1, is given by: =, = = where = =+. Now, the mobile proxy sink mps 1 moves on the outer circles of the bands b 1, b 3, b 5,,b 2m-1. That is, mps 1 moves on the circles,2 1, where = 1,2,3,...,. Thus, the energy spent by mps 1 during its mobility on the circle,2 1is given by,, = 22 1 On the other hand, mps 1 should move until band b n-1 to collect data from both bands b n-1 and b n. Thus, the energy consumed by mps 1 to move from b 1 to b n-1, denoted by,,, is given by,, = 2 1= 1 Let,, be the energy consumed by mps 1 to move from band b n-1 to band b 1, where it drops off all collected data to the sink. It is clear that mps 1 has to traverse n 1 bands (b n-1,,b 1 ), each of which has width equal to R. Thus,,, is given by,, = 1 Thus, the total energy spent per round of data collection, denoted by, is given by =,, +,, +, +,, Case 2: n is odd (n = 2m + 1) Now, the mobile proxy sink mps 1 collects data from a pair of adjacent bands during each move. However, in the last move, mps1 collects data from only one band, i.e., b n. Thus, the total average energy consumption,, which is due to the transmission of data by all source sensors located inside all bands b 1,, b n and its reception by the mobile sink mps 1, is given by: =, = = Now, mps 1 moves on the outer circles of the bands b 1, b 3, b 5,,b 2m-1. That is, mps 1 moves on the circles,2 1, where =1,2,3,...,. Thus, the energy spent by mps 1 during its mobility on the circle,2 1 is given by,, = 22 1 Notice that mps 1 should move until band b n to collect data from this band only. Thus, the energy consumed by mps 1 to move from band b 1 to band b n, denoted by,,, is given by,, = 2+1= Let,, be the energy consumed by mps 1 to move from band b n-1 to band b 1, where it drops off all collected data to the sink. It is clear that mps 1 has to traverse bands (b n,,b 1 ), each of which has width equal to R. Thus, we have,, = Therefore, the total energy spent for data collection, denoted by, is computed as follows: =,, +,, +, +,, Immobile Sink n Mobile Proxy Sinks Assuming there are n bands, our architecture requires n mobile proxy sinks, namely mps 1, mps 2, mps 3,,mps n. As stated in Theorem 2, each mobile proxy sink mps i should be moving inside its band b i along a circular trajectory of radius +, where is is the inner radius of band b i and 0< < with being the outer radius of band b i, where == 1 and =. Thus, mps i moves on the perimeter of a circle of radius 1+. Given that the bands do not have the same inner and outer radii, it is easy to check that the mobile proxy sinks located in the upper (or outer) bands would consume more energy compared to their counterparts that are located in the lower (or inner) bands. It is clear that mobile proxy sinks in the outer bands would receive more data than those located in the inner 8

9 bands, thus, yielding more energy consumption due to data reception. For instance, the number of source sensors in bands 1 and 2 are respectively given by = = = =2 =3 with >. Also, the energy spent by a mobile proxy sink due to its mobility depends on the traveled distance. As can be seen, the mobility trajectory of mobile proxy sinks in the outer bands is longer than that of mobile proxy sinks in the inner bands, thus, causing more energy consumption due to mobility. For instance, the lengths of the trajectories of the mobile proxy sinks mps 1 and mps 2 that are moving inside their respective bands b 1 and b 2, are respectively given by =2 =2 + with >. Therefore, to achieve load balancing between all n mobile proxy sinks in terms of their energy consumption, which is due to data reception and their mobility for data collection, we divide the entire circular field into n cones centered at the field center. Each cone has an angle equal to 2π/n, as shown in Figure 13. Now, each mobile proxy sink collects data from only one cone. Assuming a uniform sensor deployment, all mobile proxy sinks collect data from the same size set of source sensors. Also, these mobile proxy sinks have the same mobility trajectory as stated below. Thus, this cone-based decomposition strategy of the field ensures load balancing among all of the mobile proxy sinks in their data collection task. mobile proxy sink mps i is a sequence of arcs and segments, (epc i, v n ), [v n, v n-1 ], (v n-1, u n-1 ), [u n-1, u n-2 ], (u n- 2, v n-2 ), [v n-2, v n-3 ], (v n-3, u n-3 ),, as shown in Figure 13. After dropping off the data to the immobile sink is 1 in this first round, the mobile proxy sink mps i moves back to its original position epc i at a constant speed v max, where it pauses for some time t pause. It is clear that the sum of time spent by mps i to reach epc i and t pause are needed for the sources sensors to obtain data that will be picked up in the next round of data collection. Let us compute the total energy consumption due to data transmission by all the source sensors in a cone and their reception by the corresponding mobile proxy sink. Let be a cone centered at the circular field. Consider the portion of band b i, denoted by, which is included in the cone that is being managed by mobile proxy sink mps j. As can be seen in Figure 14, is divided into, and,. Let, and, be the number of source sensors in, and,, respectively. Then, we have:, = 1 + = 1 +2, = = 1 +2 The energy spent due to data transmission and reception to collect the data from all source sensors in is denoted by,,,, and given by:,,,,=,,, +,,, = +2,, ,, +2 Figure 13 Figure 14 Initially, each mobile proxy sink mpsi is positioned on the perimeter of the circle that divides b n into, and, at one of the two endpoints, say epc i, of its respective cone, as shown on Figure 13. For instance, in the first round, mobile proxy sink mps i follows a top-down movement from the outmost band b n until the inmost band b 1. Precisely, it moves along the arc (epc i, v n ) inside band b n, then segment [v n, v n-1 ] whose length is, then arc (v n-1, u n-1 ) inside band b n-1, etc. This mobility pattern repeats until mobile proxy sink reaches the first band b 1, where it delivers all collected data to the sink s m. Thus, the trajectory mobility of Thus, the total transmission and reception energy, denoted by,,,, which is needed to collect the data from all portions of bands, =1.., in the cone is given by:,,,=,,,, Now, let us compute the energy spent by a mobile proxy sink in one round of data collection due to its mobility. As discussed earlier, the mobility trajectory of a mobile proxy sink is a sequence of alternated segments and arcs. Each segment has length equal to except the last segment whose length is. On the 9

10 other hand, the length of an arc of a circle of radius and subtending an angle is equal to. In our case, we have = and varies between (i.e., radius of the bottom arc located in the inmost band band ) and 1+ (i.e., radius of the top arc located in the outmost band band ). The length of the arc,, denoted by,, which separates into, and, is equal to 2, = 1+ = 2 + Thus, the length of the mobility trajectory of a mobile proxy sink from band to is denoted by,,,,, and computed as,,,,,= The corresponding mobility energy, denoted by,,,,,, is given by,,,,,=,,,,, = Also, the mobile proxy sink mps j has to move up to its initial location epc i to start a new data collection round. The energy spend in this movement is denoted by,,, and given by,,,,,= 1+ Hence, the energy consumed by a mobile proxy sink during its mobility in its cone, denoted by,,,, is given by,,,= The total energy consumption in the data collection, denoted by,,,, which is due to the transmission of data by all the source sensors and its reception by mobile proxy sink mps j, is given by,,,=,,,+, = ,,,, We conclude that the average total energy consumption, denoted by, by all n mobile proxy sinks in one round of data collection is equal to,, =, =,,,+,,, 6.5 n Immobile Sinks 1 Mobile Proxy Sink Assuming there are n bands, our architecture requires n immobile sinks, namely is 1,, is n. Each immobile sink is in charge of one band and will gather the data from its respective source sensors through the single mobile proxy sink mps 1. All immobile sinks are positioned on a segment line originating from the center of the circular sensor deployment field and ending at one of the points on its perimeter. The distance between any pair of adjacent immobile sinks is equal to R, while the first immobile sink is 1 is located at a distance equal to from the center O of the circular field. In other words, every immobile sink is located at a distance equal to from the inner circle of its corresponding band. In the first round of data collection, mps 1 moves from the inmost band b 1 toward the outmost band b n. Specifically, mps 1 moves on the perimeter of a circle of radius inside band b 1 to collect data and deliver it to the immobile sink is 1. Then, it moves toward band b 2 along a segment of length R and follows a circular trajectory on the perimeter of a circle of radius + inside band b 2 to collect data from source sensors inside b 2 and drop it off to the immobile sink is 2. This pattern repeats until collecting data from all remaining bands b 3,,b n. After collecting data from band b n and delivering it to the immobile sink is n, the mobile proxy sink mps 1 pauses for some time t pause before starting its second round of data collection. This time t pause is required for all source sensors to get data. During the second round of data collection, mps 1 moves from the outmost band b n toward the inmost band b1 using the reverse path while collecting data from the source sensors in their respective bands. This mobility trajectory repeats for the network lifetime. Let,, be the average energy consumption that is due to the transmission of data by all the source sensors located inside band b i,1 and its reception by the mobile proxy sink mps 1.,, is given by,, =,, = +,, +2 = Likewise, we denote by,, the average energy consumption, which results from the transmission of data by all the source sensors located inside band b i,2 and its reception by the mobile proxy sink mps 1.,, is computed as follows:,, =,, = + +,, +2 =

11 The total average energy consumption, denoted by,, which is due to the transmission of data by all the source sensors located inside all bands b 1,, b n and its reception by the mps 1, is given by:, =,, +,, Recall that the mobility trajectory of mps 1 is a sequence of alternated circles and segments. Assuming the mobile proxy sink mps 1 is originally at the center of the field, the length of the first segment is, while the length of any subsequent segment is. Also, the radius of the first mobility circle along which mps 1 moves is, whereas the difference of the radii of two consecutive mobility circles is equal to. To summarize, when it reaches band b i, mps 1 moves on a circle of radius + 1. Thus, the energy spent by mps 1 during its mobility to collect data from all active source sensors is given by:,,= Thus, the total energy consumption,, which is necessary for the data collection process, is given by = +,, 6.6 n Immobile Sinks n mobile proxy sinks Having only one immobile sink would incur significant delay and yield considerable energy consumption due to proxy sink mobility to reach the immobile sink and deliver data to it. Assuming there are n bands, our architecture includes n immobile sinks and n mobile proxy sinks. Each band has one immobile sink and one mobile proxy sink. Each mobile proxy sink moves inside its band at a constant speed between 0 and v max along a circular trajectory to collect data from its source sensors. Then, it moves toward its immobile sink to deliver all gathered data to it. To summarize, mobile proxy sink mps i moves on the perimeter of a circle of radius 1+ inside band b i. Each immobile sink is i is positioned deterministically inside one band. Specifically, is 1 is located at the perimeter of a circle of radius inside band b 1 at location,0 while is 2 is positioned at the perimeter of a circle whose radius is + inside band b 2 at location 0, +. Then, is 3 is positioned at the perimeter of a circle of radius +2 inside band b 3 at location 2,0 whereas is 4 is located at the perimeter of a circle with radius +3 inside band b 4 at location 0, 3. Then, is 5 is positioned at the perimeter of a circle of radius +4 inside band b 5 at location +4,0, etc. To summarize, the immobile sink is i is located at the perimeter of a circle of radius + 1 in band b i at location (x, y), where 1 and x and y are computed as follows:,= + 1,0 if i mod 4 = 1,=0, + 1 if i mod 4 = 2,= 1,0 if i mod 4 = 3,=0, 1 if i mod 4 = 0 It is worth mentioning that we propose such a placement of immobile sinks to avoid interference and collisions due to simultaneous transmission of data by source sensors to their respective mobile proxy sinks. Let,, be the average energy consumption that is due to the transmission of data by all the source sensors in band b i,1 and its reception by mobile proxy sink mps i. This energy,, is computed as:,, =,,, = +,, +2 = Let,, be the average energy consumption that results from the transmission of data by all the source sensors located inside band b i,2 and its reception by the mobile sink mps i. This energy,, is given by,, =,,, = = The total average energy consumption,, due to the data transmission by all the source sensors in band b i and its reception by mps i, is given by:, =,, +,, Also, the energy spent by mps i during its mobility along the perimeter of a circle of radius + 1 inside band b i is given by,,,= Thus, the average total energy, denoted by, which is spent by all the source sensors and mobile proxy sinks in the data collection process is =, +,,, 7. SIMULATION RESULTS In this section, we present the simulation results of our data collection protocols using a high-level simulator written in C language. First, we specify our simulation environment. Then, we discuss our simulation results. 11

12 7.1 Simulation Environment We consider a circular deployment field of radius D = 1000 m. We assume that the radii of the sensing and communication ranges of all the sensors (including both of the source sensors and mobile proxy sinks) are r = 50 m and R = 100 m, respectively. Thus, the number of concentric circular bands is n = D/R = 10. Moreover, we consider the free-space model, where =2. In addition, according to the energy model given in [28], the energy consumption in transmission, reception, idle, and sleep modes are 60 mw, 12 mw, 12 mw, and 0.03 mw, respectively [28]. We suppose that the initial energy of each source sensor is 60 Joules, while that of each mobile proxy sink is 100 Joules. To the best of our knowledge, there is no experimental result that has reported in the literature regarding the energy e move that is spent per unit of distance due to sensor mobility. In the absence of such an empirical value, we assign two different values to e move to assess the impact of mobility. We assume that the moving speed of each proxy sink is constant and equal to 1m/s. All simulations are repeated 100 times and the results are averaged. Figure 17: Protocol 1 n Figure 18: Protocol n 1 Figure 15: Protocol 1 1 (Scenario 1) Figure 16: Protocol 1 1 (Scenario 2) Figure 19: Protocol n n 7.2 Simulation Results We use the distributed connected k-coverage protocol, called DIRACC k, which is proposed for homogeneous WSNs, where all the source sensors and the single immobile sink are immobile [1]. Here, we focus on the energy consumption due to the data collection process (data transmission, data reception, and mobility). 12

13 In all the experiments related to Figures 15 19, we set the value of to, i.e., = = Let our data collection protocols described in Sections 6.3, 6.4, 6.5, and 6.6, be called 1 1 protocol, 1 n protocol, n 1 protocol, and n n protocol, respectively. Figure 15 shows the results of the first scenarios of the 1 1 protocol, while Figure 16 shows those of the second scenario. We found a match between theoretical and simulation results. Also, the 1 1 protocol in Scenario 1 gives better results than in Scenario 2. This is due to the different transmission distances used by the source sensors. In Scenario 1, the average transmission distances are,, = /3 (in band, ) and,, = /3 (in band, ), while in Scenario 2, it is equal to,, =/3. Figure 17 shows the theoretical and simulation results of the 1 n protocol. Likewise, there is a match between both results. In Figure 18, we plot both of the theoretical and simulation results of the n 1 protocol. Similarly, we notice that there is a match between both results. Figure 19 shows the theoretical and simulation results of the n n protocol, with a good match between them. Comparing Figure 17 and Figure 18, it is easy to see that the n 1 protocol outperforms the 1 n protocol. While they both have the same energy consumption due to data transmission and reception, the 1 n protocol incurs more mobility given that is has n mobile proxy sinks. Indeed, each of the n mobile proxy sinks has to move towards the center of the field to deliver the collected data to the single static sink. Thus, the 1 n protocol is expected to consume more energy than the n 1 protocol. We found that both of the n 1 protocol and n n protocol have comparable energy consumption. We expect the n n protocol to consume the least amount of energy for data collection. Indeed, compared to the n 1 protocol, there is no rectilinear movement in the n n protocol as every mobile proxy sink is in charge of only one band. Also, there is no need for any mobile proxy sink in the n n protocol to move out of its band as there is an immobile sink in its corresponding band to deliver data to. Thus, the energy savings of the n n protocol is due to the absence of rectilinear mobility of mobile proxy sinks. Now, we consider a different value of in the experiments related to Figures 20 21, where = Figure 20 shows the energy consumption during data collection and delivery of the 1 1 protocol, while Figure 21 shows that of the of the 1 n protocol. It is clear that the former outperforms the latter. In fact, the 1 n protocol requires more mobility given that all n mobile proxy sinks have to move toward the center of the field to deliver the collected data to the static sink. Thus, the selection of the best protocol with regard to the number of mobile proxy sinks to solve the energy sink-hole problem should depend on the mobility cost. Figure 20: Protocol 1 1 (e move = J) Figure 21: Protocol 1 n (e move = J) 8. CONCLUSION In this paper, we investigated the problem of energy sink-hole in duty-cycled connected k-covered WSNs. First, we proposed a three-tier architecture that has immobile source sensors, immobile sinks, and mobile proxy sinks. Then, we computed the optimum proxy sink mobility strategy with a goal to minimize the total energy consumption due to data transmission by source sensors and its reception by a mobile proxy sink. Also, we proposed four data collection protocols based on the number of immobile sinks and mobile proxy sinks. Second, we provided a theoretical analysis of each of those data collection protocols. Third, we corroborated our analysis with several simulation results. We found a good match between analytical and simulation results. We also show that our data collection protocols outperform a solution not using mobile proxy sinks. Our future work is three-fold. First, we plan to study delay in our proposed data collection protocols. Second, we focus on the problem of energy sink-hole problem in sparsely deployed WSNs. Third, we plan to extend our work to three-dimensional WSNs. 13

14 REFERENCES [1] H. M. Ammari and S. K. Das, Centralized and clustered k-coverage protocols for wireless sensor networks, IEEE TC, 61(1), [2] X. Bai, S. Kumar, D. Xuan, Z. Yun, and T. Lai, Deploying wireless sensors to achieve both coverage and connectivity, Proc. ACM MobiHoc, [3] B. Bollobás, The Art of Mathematics: Coffee Time in Memphis, Cambridge Univ. Press, [4] N. Bulusu, J. Heidemann, and D. Estrin, GPSless low cost outdoor localization for very small devices, IEEE PCM, 7(5), [5] E. Efthymiou, S. Nikoletseas, and J. Rolim, Energy balanced data propagation in wireless sensor networks, WINET, 12, [6] Y. Ganjali and A. Keshavarzian, Load balancing in ad hoc networks: single-path routing vs. multipath routing, Proc. IEEE Infocom, [7] J. Gao and L. Zhang, Load balanced short path routing in wireless networks, Proc. IEEE Infocom, [8] W. Guo and G. Wu, An energy-balanced transmission scheme for sensor networks, Proc. ACM SenSys, [9] H. Gupta, Z. Zhou, S. Das, and Q. Gu, Connected sensor cover: Self-organization of sensor networks for efficient query execution, IEEE TON, 14, [10] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, An application-specific protocol architecture for wireless microsensor networks, IEEE TWC, 1(4), [11] C. Huang, Y. Tseng, and H. Wu, Distributed protocols for ensuring both coverage and connectivity of a wireless sensor network, ACM TOSN, 3(1), [12] A. Jarry, P. Leone, O. Powell, and J. Rolim, An optimal data propagation algorithm for maximizing the lifespan of sensor networks, Proc. DCOSS, LNCS 4026, [13] L. A. Klein, A boolean algebra approach to multiple sensor voting fusion, IEEE TAES, 29(2), [14] P. Leone, S. Nikoletseas, and J. Rolim, An adaptive blind algorithm for energy balanced data propagation in wireless networks, Proc. DCOSS, LNCS 3560, [15] J. Li and P. Mohapatra, Analytical modeling and mitigation techniques for the energy hole problem in sensor networks, Elsevier PMC, 3, [16] J. Lian, K. Naik, and G. Agnew, Data capacity improvement of wireless sensor networks using non-uniform sensor distribution, IJDSN, 2-2, [17] J. Luo and J.-P. Hubaux, Joint mobility and routing for lifetime elongation in wireless sensor networks, Proc. IEEE Infocom, [18] V. P. Mhatre, C. Rosenberg, D. Kofman, R. Mazumdar, and N. Shroff, A minimum cost heterogeneous sensor network with a lifetime constraint, IEEE TMC, 4(1), [19] D. Nicules and B. Nath, Ad-hoc positioning system using AoA, Proc. IEEE Infocom, [20] S. Olariu and I. Stojmenovic, Design guidelines for maximizing lifetime and avoiding energy holes in sensor networks with uniform distribution and uniform reporting, Proc. IEEE Infocom, [21] O. Powell, P. Leone, and J. Rolim, Energy optimal data propagation in wireless sensor networks, Journal of Parallel and Distributed Computing, 3(67), pp , Mar [22] T. Sun, L. Chen, C. Han, and M. Gerla, Reliable sensor networks for planet exploration Proc. IEEE NSC, [23] Y.-C. Wang and Y.-C.Tseng, Distributed deployment schemes for mobile wireless sensor networks to ensure multi-level coverage, IEEE TPDS, 19(9), [24] X. Wu, G. Chen, and S. K. Das, Avoiding energy holes in wireless sensor networks with nonuniform node distribution, IEEE TPDS, 19(5), [25] G. Xing, X. Wang, Y. Zhang, C. Lu, R. Pless, and C. Gill, Integrated coverage and connectivity configuration for energy conservation in sensor networks, ACM TOSN, 1(1), pp , [26] S. Yang, F. Dai, M. Cardei, and J. Wu, On connected multiple point coverage in wireless sensor networks, IJWIN, 13(4), [27] M. Yarvis, N. Kushalnagar, H. Singh, A. Rangarajan, Y. Liu, and S. Singh, Exploiting heterogeneity in sensor networks, Proc. IEEE Infocom, [28] F. Ye, G. Zhong, J. Cheng, S. Lu, and L. Zhang, PEAS: A robust energy conserving protocol for long-lived sensor networks, Proc. ICDCS, [29] H. Zhang and J. Hou, Maintaining sensing coverage and connectivity in large sensor networks, AHSWN, 1(1-2), pp , [30] H. Zhang, H. Shen, and Y. Tan, Optimal energy balanced data gathering in wireless sensor networks, Proc. IEEE IPDPS, [31] Y. Zou and K. Chakrabarty, A distributed coverage- and connectivity-centric technique for selecting active nodes in wireless sensor networks, IEEE TC, 54-1,

15 Habib M. Ammari is an Associate Professor in the Department of Computer and Information Science, College of Engineering and Computer Science, University of Michigan-Dearborn, and the Founding Director of Wireless Sensor and Mobile Ad-hoc Networks (WiSeMAN) Research Lab at the University of Michigan-Dearborn since September Prior to that, he was on the faculty of the Department of Computer Science, Hofstra University from September August He obtained his second Ph.D. degree in Computer Science and Engineering from the University of Texas at Arlington in May 2008, and his second Master's degree in Computer Science from Southern Methodist University in December Also, he obtained his first Ph.D. (Highly Honorable with Praise) and Master's degrees in Computer Science from the Faculty of Sciences of Tunis in December 1996 and July 1992, respectively. His main research interests lie in the areas of wireless sensor and mobile ad hoc networking, multi-hop mobile wireless Internet architectures and protocols, and cyber physical systems. In particular, he is interested in coverage, connectivity, energy-efficient data routing and information dissemination, fault tolerance, and security in wireless sensor networks, and the interconnection between wireless sensor networks, mobile ad hoc networks, and the global IP Internet. He has a strong publication record in top-quality journals, such as IEEE TPDS, IEEE TC, ACM TAAS, Elsevier COMNET, Elsevier PMC, Elsevier JPDC, and highquality conferences, such as IEEE SECON, IEEE ICDCS, EWSN, and ICDCN. He published his first Springer book, titled Challenges and Opportunities of Connected k-covered Wireless Sensor Networks: From Sensor Deployment to Data Gathering, in August He received several prestigious awards, including the Certificate of Appreciation Award at the 17 th ACM MobiCom in September 2011, the Outstanding Leadership Award at the 20 th IEEE ICCCN in August 2011, the Best Symposium Award at the 7 th IEEE IWCMC in July 2011, the Lawrence A. Stessin Prize for Outstanding Scholarly Publication from Hofstra University in May 2010, the Faculty Research and Development Grant Award from Hofstra College of Liberal Arts and Sciences in May 2009, the Best Paper Award at the 5 th EWSN in February 2008, and the Best Paper Award at the Google Ph.D. Forum 6 th IEEE PerCom in March Also, he was an ACM Student Research Competition (ACM SRC) nominee at ACM MobiCom He is the recipient of the Nortel Outstanding CSE Doctoral Dissertation Award in February 2009, and the John Steven Schuchman Award for Outstanding Research by a PhD Student in February He received a three-year US National Science Foundation (NSF) Research Grant Award ($400 K) in June 2009, and the US National Science Foundation Faculty Early Career Development (CAREER) Award ($450 K) in January He is the University of Michigan- Dearborn s second NSF CAREER Award recipient, and the Hofstra's first ever NSF CAREER Award recipient. He serves as Associate Editor of several journals, including ACM TOSN, IEEE TC, Wiley WCMC, AHSWN, Wiley IJCS, and NPA. He is on the Editorial board of the International Journal of Mobile Communications and the International Journal on Advances in Networks and Services. Also, he is on the Editorial Review Board of the International Journal of Distributed Systems and Technologies. He serves as Track Co-Chair of IEEE ICCCN 2012 SEP, Demo Co-Chair of IEEE WoWMoM 2012, Program Co- Chair of CPNS 2012, Program Vice-Chair of MUSIC 2012, Workshop Chair of WiMAN 2012, Program Vice-Chair of IEEE/IFIP EUC 2011, Workshop Chair of WiMAN 2011, and Symposium Co-Chair of IWCMC 2011 Wireless Sensor Networks Symposium. Also, he served as Workshop Co-Chair of WiMAN 2010, Symposium Co-Chair of IWCMC 2010 Wireless Sensor Networks Symposium, Program Co-Chair of IQ2S 2009, and Workshop Co-Chair of WiMAN He has served as Publicity (Co) Chair of numerous conferences, symposia, and workshops, including ACM MobiCom Also, he has served as a reviewer for several international journals, including IEEE TMC, IEEE TPDS, IEEE SMC, ACM TOSN, IEEE TVT, IEEE TWireless, IEEE CL, Springer s MONE, Springer s WINET, Elsevier s ADHOC, Elsevier s COMNET, AHSWN, IJSNet, Elsevier s IPL, Elsevier s COMCOM, Elsevier s JPDC, Elsevier s INS, IJCA, and Elsevier s DKE, and as a Technical Program Committee member of numerous IEEE and ACM conferences, symposia, and workshops, including IEEE Infocom, IEEE ICDCS, IEEE DCOSS, IEEE INSS, IEEE PerCom, IEEE GlobeCom, IEEE ICC, SSS, IEEE MASS, IEEE MSN, IEEE LCN, IEEE VTC, IEEE ICCCN, EWSN, ICDCN, and AdHocNets. 15