Crosslayer optimized routing in wireless sensor networks with duty ...

WIRELESS COMMUNICATIONS AND MOBILE COMPUTING Wirel. Commun. Mob. Comput. (2014) Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/wcm.2468

RESEARCH ARTICLE

Cross-layer optimized routing in wireless sensor networks with duty cycle and energy harvesting Guangjie Han1,2 , Yuhui Dong3 , Hui Guo1 , Lei Shu2* and Dapeng Wu4 1

Department of Information & Communication Engineering, Hohai University and Changzhou Key Lab. of Photovoltaic System Integration and Production Equipment Technology, Changzhou, China 2 Guangdong Provincial Key Lab. of Petrochemical Equipment Fault Diagnosis, Guangdong University of Petrochemical Technology, China 3 The Patent Examination Cooperation Center of SIPO, Jiangsu, China 4 Department of Electrical and Computer Engineering, University of Florida, Gainesville, FL, U.S.A.

ABSTRACT In this paper, we propose a cross-layer optimized geographic node-disjoint multipath routing algorithm, that is, twophase geographic greedy forwarding plus. To optimize the system as a whole, our algorithm is designed on the basis of multiple layers’ interactions, taking into account the following. First is the physical layer, where sensor nodes are developed to scavenge the energy from environment, that is, node rechargeable operation (a kind of idle charging process to nodes). Each node can adjust its transmission power depending on its current energy level (the main object for nodes with energy harvesting is to avoid the routing hole when implementing the routing algorithm). Second is the sleep scheduling layer, where an energy-balanced sleep scheduling scheme, that is, duty cycle (a kind of node sleep schedule that aims at putting the idle listening nodes in the network into sleep state such that the nodes will be awake only when they are needed), and energy-consumption-based connected k-neighborhood is applied to allow sensor nodes to have enough time to recharge energy, which takes nodes’ current energy level as the parameter to dynamically schedule nodes to be active or asleep. Third is the routing layer, in which a forwarding node chooses the next-hop node based on 2-hop neighbor information rather than 1-hop. Performance of two-phase geographic greedy forwarding plus algorithm is evaluated under three different forwarding policies, to meet different application requirements. Our extensive simulations show that by cross-layer optimization, more shorter paths are found, resulting in shorter average path length, yet without causing much energy consumption. On top of these, a considerable increase of the network sleep rate is achieved. Copyright © 2014 John Wiley & Sons, Ltd. KEYWORDS cross-layer optimization; geographic multipath routing; 2-hop; energy harvesting *Correspondence Lei Shu, Guangdong Provincial Key Lab. of Petrochemical Equipment Fault Diagnosis, Guangdong University of Petrochemical Technology, Guangzhou, China. E-mail: [email protected]

1. INTRODUCTION Recently, sensors are developed in wireless sensor networks (WSNs) to scavenge energy from the natural environment, such as solar, heat, or vibration [1–3]. Although these sensors may be alive again after replenishing the energy, their temporary failure may still affect the quality of service provided by WSNs, for example, the temporally dead status of a critical node may result in the partition of the whole network [4]. Thus, sleep scheduling should be still applied in rechargeable WSNs, not only reducing their energy consumptions when the energy harvesting is not very efficient but also leaving sensor nodes enough time to recharge the energy. Copyright © 2014 John Wiley & Sons, Ltd.

In terms of data dissemination in large-scale WSNs, geographic routing is commonly adopted, in which packets are forwarded locally and greedily to the 1-hop neighbor closest to the destination [5,6]. A variety of other forwarding mechanisms have also been researched. GeRaF [7] considers a geographical forwarding in a wireless mesh network, which is sleep–wake cycling to a sleep–wake cycling for save energy. Ammari et al. [6] also propose a design of a geographic forwarding protocol for duty-cycled k-covered WSNs with data aggregation. In our previous work [8], we conducted the performance analysis on the connected k-neighborhood sleep scheduling algorithm (CKN)-based [9] duty-cycled WSNs for TPGF [10] nodedisjoint multipath routing algorithm and show the fact that

G. Han et al.

Optimized geographic node-disjoint multipath routing algorithm

it is possible to keep a portion of sensor nodes in sleep mode and still provide a certain level quality of service for WSN applications. Ruehrup et al. [11] propose a novel online routing scheme to provide loop-free, fully stateless, energy-efficient sensor-to-sink routing at a low communication overhead without the help of prior neighborhood knowledge. While existing researches mostly focus on 1hop geographic forwarding, Stojmenovic and Lin [12] have extended existing geographic routing schemes by considering selecting the next-hop node from 2-hop neighbors. It has been shown that this 2-hop routing algorithm increases the success rate compared with the 1-hop variant. However, most of the aforementioned work have been solely concerned with the design of different forwarding mechanisms for performance improvement. Few give comprehensive considerations to energy replenishment techniques, adjustable transmission power, and sleep scheduling strategy, that is, cross-layer approaches. Cross-layer approaches attempt to exploit a richer interaction among communication layers to achieve performance gains. For instance, both medium access and routing decisions have significant impact on power consumption, and the joint consideration of both can yield more efficient power consumption [13]. EYESMAC [14] has modeled the interaction between the MAC and routing protocol. It can improve traffic routing performance with consideration of network topologies, power duty cycling, and node failure. In [15], MACRO integrates MAC and routing layer functionalities to support geographic forwarding in WSNs, which adjusts the transmission power and selects the best next relay node while forwarding information to the destination. In this paper, we seek a cross-layer optimized routing algorithm. Physical layer, sleep scheduling layer, and routing layer work together to choose the optimized transmission route. More specifically, we introduce an energy harvesting model, and each node is able to adjust its transmission power depending on its current energy level. Moreover, we apply an energy-balanced sleep scheduling scheme (EC-CKN) [16] for our sleep scheduling layer. The physical layer provides the remaining energy and transmission radius information for the sleep scheduling layer, which dynamically schedules sleep rate of the network. Then, the routing layer chooses those routes that reduce energy consumption, and the physical layer adjusts the transmission power accordingly. In short, we will study several important issues. First, how to find a good approach to facilitate the geographic routing in duty-cycled WSNs with energy harvesting? Cross-layer approaches are promising solutions [17]. An optimization framework is needed to concurrently model multiple parameters from equivalent layers. Particularly, we should solve the uncertainty of the remaining energy in sensor nodes and try to always select the sensor nodes with more remaining energy to perform the data dissemination task. Second, whether the existing geographic forwarding policy with 1-hop neighbor information is suitable for duty-cycled WSNs? Further discussion is necessary. Finally, how to design an efficient geographic node-disjoint

multipath routing algorithm that allows higher sleep rate in the network which can be implemented in WSNs? In addressing the aforementioned issues, we propose a novel routing algorithm TPGFPlus. To the best of our knowledge, TPGFPlus is the first cross-layer optimized work to consider 2-hop based geographic routing for dutycycled and energy renewable WSNs, which is inherently loop free. We also introduce three different forwarding policies and study their distinct performance. On the basis of extensive simulations, we find that geographic routing in duty-cycled WSNs should be 2-hop based, but not 1-hop based. For these 2-hop-based geographic forwarding policies, all of them achieve better routing performance than the 1-hop one (TPGF), in terms of both the average number of explored paths and the average path length. What is more, our cross-layer optimized routing allows considerable higher sleep rate in the network. The rest of this paper is organized as follows. Section 2 reviews related work. Section 3 introduces the cross-layer optimized framework of our research work. Section 4 describes the design of our algorithm. Section 5 presents properties of TPGFPlus, and Section 6 is related to performance study. Finally, we conclude this paper in Section 7.

2. RELATED WORK 2.1. Energy harvesting and adjustable transmission power Recently, WSNs are equipped with solar power cells to solve the original shortcoming that the network lifetime is limited by using the batteries [1]. Voigt et al. [2] design two solar-aware routing protocols that preferably route packets via solar-powered nodes and show that routing protocols provide significant energy savings. Kai Zeng et al. propose two protocols, GREES-L and GREES-M, which make routing decision locally by jointly taking into account the realistic wireless channel condition, packet progress to the destination, the residual battery energy level of the node, and the environmental energy supply [5]. Stojmenovic et al. [18] discuss the case of adjustable transmission radii for geographic routing. This work is an example of a more complex and intelligent network layer that takes into account lower layer parameters together with a more realistic modeling of lower layers. In [19], nodes do not have to transmit packets with their maximum transmission ranges all the time. They take their energy resource into consideration and adjust their transmission range accordingly. 2.2. Connected k -neighborhood and energy-consumption-based connected k -neighborhood sleep scheduling algorithms Sensors in WSNs are normally duty cycle-based for periodically sleeping to save energy and prolong network Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


lifetime. Generally, duty cycle in WSNs can be classified into two basic categories: random duty cycle and scheduled duty cycle. The major difference between random duty-cycled WSNs and scheduled duty-cycled WSNs is the time-varying network connectivity. First, in random duty-cycled WSNs, the network-wide connectivity is not guaranteed. Second, in scheduled duty-cycled WSNs, the network-wide connectivity is guaranteed. In this paper, we mainly consider the scheduled duty-cycled WSNs, for example, CKN-applied [9] and EC-CKN-applied [16] WSNs. The CKN algorithm allows a portion of sensor nodes to go to sleep but still keeps all awoken sensor nodes kconnected to prolong the lifetime of a WSN. Here, the parameter k represents the required minimum number of awake neighbors per node. These requirements ensure that if a node has less than k neighbors, none of its neighbors goes to sleep, and if it has more than k neighbors, at least k neighbors of them decide to remain awake [16]. Then, geographic routing can be carried on duty-cycled nodes. This algorithm provides the first formal analysis of the performance of geographic routing on duty-cycled WSNs, where every sensor has k awake neighbors. A variant of this method called EC-CKN prolongs network lifetime further. Different from CKN, EC-CKN takes nodes’ residual energy information as the parameter to schedule nodes to be active or sleep. It makes sure that awake neighbor nodes have more residual energy than other neighbor nodes at each epoch. It is worth noting that the uncertainty of remaining energy level that caused by both environmental energy harvesting and package transmission is handled by the EC-CKN algorithm. 2.3. Geographic multipath routing in duty-cycled wireless sensor networks Although many multipath routing protocols have been studied [20,21], most of them focus on reducing delay, providing reliability, reducing overhead, maximizing network lifetime, or supporting hybrid routing and are extended versions of Dynamic Source Routing (DSR) and Ad hoc On-demand Distance Vector (AODV) routing. These routing protocols cannot provide a powerful searching mechanism for the maximum number of shortest paths, as well as bypassing holes [22], not to mention in duty-cycled WSNs. For such drawbacks, we propose TPGFPlus, which is a geographic node-disjoint multipath routing algorithm in duty-cycled WSNs. Moreover, there is no existing multipath routing focusing on exploring geographic multipath routing in dutycycled WSNs with energy harvesting. 2.4. Geographic routing with 2-hop neighbors Geographic routing uses position information to forward a message to its destination. Many research works have been Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

performed to obtain the position information [23–25]. Sensor nodes are not required to maintain global and detailed information on the entire network topology. They only need to maintain local knowledge on their 1-hop neighborhood information with respect to their final destination. Although most geographic routing protocols use 1-hop information, generalization to 2-hop neighborhood is also possible [12]. Extending geographic routing schemes to 2-hop neighborhood increases success delivery rate, referring to the geographic routing based on 1-hop neighborhood. Another algorithm called normalized advance (NADV) [26] also considers a geographic routing scheme that uses 2-hop neighborhood information. Not surprisingly, utilizing 2-hop neighborhood information leads to higher-quality paths than the 1-hop case. However, these schemes still cannot guarantee that packets are delivered in an energy-efficient manner. In real networks, nodes commonly employ low duty-cycling and are asleep at most time to have a long battery life [27]. This important feature is ignored in previous work. 2.5. Two-phase geographic greedy forwarding geographic multipath routing Our previous work, TPGF [10] is one of the earliest researches on geographic node-disjoint multipath routing with 1-hop neighborhood in wireless multimedia sensor networks. TPGF does not require the computation and preservation of the planar graph in WSNs. This point allows more links to be available for TPGF to explore more node-disjoint routing paths, because using the planarization algorithms actually limits the useable links for exploring possible routing paths. Greedy Distributed Spanning Tree Routing (GDSTR) [28] is another geographic routing approach that does not apply the planar graph but maintains two or more hull trees. However, greedy distributed spanning tree routing does not support multipath routing, because the number of available links in the hull trees topology is limited. Our TPGF algorithm includes two phases: Phase 1 is responsible for exploring a delivery guaranteed routing path while bypassing holes. Phase 2 is responsible for optimizing the found routing path with the least number of hops. Two-phase geographic greedy forwarding algorithm finds one path per execution and can be executed repeatedly to find more node-disjoint routing paths with the guarantee that any node will not be used twice. It makes three practical contributions: (i) supporting multipath transmission; (ii) supporting hole-bypassing, not only static holes but also dynamic holes that may occur if several sensors in a small area overload due to data transmission; and (iii) supporting shortest path transmission.

G. Han et al.


2.6. Different forwarding polices for geographic routing Geographic routing is characterized by the greedy forwarding policies applied to each routing step, which result in different routing paths. There have been extensive researches on choosing routing policies for different applications and network assumptions, for example, considering the energy balancing [29] or link quality [30]. Li et al. [29] propose a local power efficiency metric for geographic routing such that at each step the transmitter picks as the next hop neighbor for which this metric is maximized. Seada et al. [30] articulate the distance-hop energy trade-off for geographic routing and conclude that packet reception rate . Distance is an optimal local link metric for making geographic routing decisions in lossy wireless networks. All these approaches mostly make forwarding decisions based on different metrics, but few simultaneously consider the potential of harvesting energy from environment sources except [5]. However, they do not consider sleep/awake cycles.

3. THE CROSS-LAYER OPTIMIZED FRAMEWORK In this section, to better describe the interactions of the multiple layers, we present a cross-layer optimized framework underlying our proposed TPGFPlus scheme. It takes into account the first layer, physical layer, where sensor nodes are developed to scavenge energy from environment, and each node can adjust its transmission power depending on its current energy level; the second layer, sleep scheduling layer, where an energy balanced sleep scheduling scheme is applied to allow sensor nodes to get enough time to recharge energy, which takes nodes?current energy level as the parameter to dynamically schedule nodes to be active or asleep; and the third layer, routing layer, in which a forwarding node chooses the next-hop node based on 2-hop neighbor information rather than 1-hop. Further, performance of TPGFPlus protocol will be evaluated under three different forwarding policies, to meet different application requirements. As illustrated in Figure 1, our framework allows interactions among different layers to optimize system performance as a whole. 3.1. Layer 1: physical layer Now, we define the network model used throughout this paper. First, we introduce a basic definition. 2-hop neighbors. Let N.vi / and N.vi /0 be respectively the sets of node vi ’s 1-hop and 2-hop neighbor nodes, that is, vi ’s 2-hop neighbors are the neighbors of vi ’s 1-hop neighbors after removing the duplicated ones. Here, we mainly consider the simplest scenario with only one source node in the outdoor network. We model a network with N sensor nodes randomly deployed. The locations of sensor nodes and the base station are fixed

and can be obtained by using GPS. Each node knows its own location and the position information of its 1-hop and 2-hop neighbors. The Euclidean distance between any two node vi and vj is denoted as Disti,j . Both the source node and the sink node are assumed with unlimited power supply. Each normal sensor node is powered by rechargeable batteries with the capability to harvest solar power to recharge the batteries from one additional solar power cell. The energy harvesting rate and energy consumption rate in each individual sensor node are different and unpredictable, because many factors can contribute to this and affect the residential energy level, for example, the unstable local weather, the different number of 1-hop neighbor nodes, and the unexpected query tasks from users. Also, we consider the case when each node uses an adjustable transmission power depending on their battery level. Each node has a fixed number of transmission power levels. An example of such sensor nodes are Berkeley Motes [31]. Instead of transmitting at maximum power, nodes take their energy resource into consideration and collaboratively adjust their transmission power accordingly [19]. Any node, if rich in residual energy, will have the ability of enlarging its maximum transmission radius; thus, network connectivity and network sleep rate are directly affected.

3.2. Layer 2: sleep scheduling layer In the sleep scheduling layer, to balance energy consumption and prolong network lifetime, we assume all nodes are operating with EC-CKN-based sleep/awake duty-cycling. The 2-hop neighbors are gathered when executing ECCKN for sleep scheduling in WSNs. Each sensor dynamically turns on and off the radio in turn based on the 2-hop neighbors remaining energy information. Time is divided into epoches, and each epoch is represented by T. In each epoch, the node will first transmit packets and then run the EC-CKN sleep/awake scheduling algorithm to schedule the state of the next epoch: sleep or awake. Sensor nodes takes their current energy level information as the parameter to decide whether a node to be active or sleep, dynamically adjust the sleep rate of network. The parameter k, which is considered as a method for controlling the sleep rate of the network, can directly affect the number of awoke nodes for geographic routing.

3.3. Layer 3: routing layer In this layer, we propose a 2-hop geographic multipath routing named TPGFPlus, which we will introduce in detail in next section. As shown in Figure 1, our protocol is cross-layer optimized algorithm based on joint consideration of different underlying layers. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


Figure 1. The cross-layer optimized framework.

4. DESIGN OF TWO-PHASE GEOGRAPHIC GREEDY FORWARDING PLUS ALGORITHM In our routing algorithm, each node locally maintains its 1-hop and 2-hop neighbors’ information such as location, residual energy, energy harvested rate, and energy consumed rate. We implement 2-hop geographic forwarding for a cross-layer optimized multipath routing. The gathering of 2-hop neighbors is not an additional overhead for TPGFPlus algorithm, because the 2-hop neighborhood information is obligatorily gathered when executing ECCKN. But extending to 3-hop or even more will incur extra

broadcasting, which is not included in EC-CKN any more. In summary, TPGFPlus algorithm consists of two phases: (i) 2-hop geographic forwarding and (ii) path optimization. 4.1. 2-hop Geographic forwarding In this phase, we introduce two courses: greedy forwarding and step back & mark. The policy for greedy forwarding in this paper is as follows. Suppose a current forwarding node always chooses its next-hop node which is closest to the based station among all its 1-hop and 2-hop neighbor nodes and the nexthop node to the base station can be further than itself. Once

Figure 2. TPGFPlus 2-hop geographic forwarding example: node a chooses a 2-hop neighbor node g as its next-hop node that is closest to the sink among all a’s 1-hop and 2-hop neighbors. Once node g is chosen, node a must first select node b as an intermediate 1-hop direct neighbor based on a certain selecting policy. As shown in Figure 2, if node g has a considerable higher energy level, it can enlarge its transmission power. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


ALGORITHM 1.TPGFPlus Algorithm (* Run the following at each node vi *) 1. When a node vi receives a packet for the sink, it checks whether the sink is in its 1-hop neighbor set N.vi /; if(sink 2 N.vi /), forward the packet to the sink; else goto step 2; 2. Acquire the geographic distance drankvi to the sink for each non-blocked node of its 1-hop and 2-hop neighbor sets, N.vi / and N.vi /0 . Let Dvi be the set of these ranks. A degressive number based-label Nid will be given to each chosen next-hop node along with a path number Pid . 3. Choose node vj with minimum value of drankvi among all remaining neighbors. 4. Check whether next-hop node vj is already on the path, which may cause path circle; if(vj is already on the path), goto step 3; else if(vj 2 N.vi /), directly forward the packet to the chosen node vj ; else if(vj 2 N.vi /0 ), must first find an intermediate forwarding 1-hop neighbor vk , according to some policy. Through this chosen node vk , the packet is transmitted to node vj ; 5. Let N.vi /resi be the set of non-blocked 1-hop neighbors. if(N.vi /resi D ;), step back to the previous hop node; make itself as a block node; goto step 1; 6. The algorithm will terminate when the packet is forwarded to the sink or there are no non-blocked neighbors of source node remaining. if(the routing path is built up), a successful ACK is sent back from the base station to the source node. Any node in the path relays ACK to its 1hop neighbor node with the same Pid and largest Nid ; else an unsuccessful ACK is created when the path exploration reach the source node, which means no any other path can be found.

the forwarding node chooses its next-hop node among its 2-hop neighbor nodes that have not been labeled, it will have to find an intermediate 1-hop direct neighbor that has not been labeled according to some selecting policy. A digressive number-based label is given to the chosen sensor node along with a path number, which will be kept during the path exploration time, but not always. Thus, it is feasible even for resource constrained sensors. This greedy forwarding principle is different from the greedy forwarding principle in [32–35]: a forwarding node always chooses the 1-hop neighbor node that is closer to the base station than itself. And, the local minimum problem does not exist. Supposing candidate nodes with similar progress to the destination, the one with higher energy harvesting rate

and residual energy will be chosen first. In this point, our work is significantly different from the previous TPGF [10] where the strategy is forwarding the packets to the direct 1hop neighbor which is nearest to the sink. Figure 2 simply describes the geographic forwarding process of TPGFPlus. The key difference between with 1-hop and 2-hop, actually in 1-hop routing, is that the forwarding strategy is always forwarding packets to the direct 1-hop neighbor that is nearest to the sink. But in 2-hop routing, a current forwarding node always chooses its next-hop node that is closest to the based station among all its 1-hop and 2-hop neighbor nodesïijZ˙ once the forwarding node chooses its next-hop node among its 2-hop neighbor nodes that have not been labeled, it will have to find an intermediate 1-hop direct neighbor that has not been labeled according to some selecting policy. Although such a method does not have well-known local maximum problem, there may be block situations [10]. During the discovering of a path, if any forwarding node has no 1-hop neighbors except its previous-hop node, we will mark this node as a block node and this situation as a block situation. In this situation, the step back & mark course will start. The block node will step back to its previous-hop node, which will attempt to find another available neighbor as next-hop node. This course will be repeatedly executed until a node successfully finds a next-hop node to convert back to greedy forwarding course. Observe that TPGFPlus does not include the face routing concept, for example, right/left hand rules and count/clockwise angles, does not require the computation and preservation of the planar graph and does not have the well-known local minimum problem, making it different from existing geographic routing algorithms. 4.2. Path optimization Path circle. As shown in Figure 3(a), for any given routing path, if two or more than two nodes in a path are neighbor nodes of another node in this path, we consider that there is a path circle. To optimize the found routing path, we eliminate unnecessary circles in it. Although our work uses 2-hop neighborhood information, path circles also appear. To eliminate the path circles and optimize the found routing path with the least number of hops, we introduce the label-based optimization. The principle of this label-based optimization is that any node in a path only relays the acknowledgement to its 1hop neighbor node that has the same path number and the largest node number. In Figure 3(b), after path optimization, we get a shorter path. Once the optimized path is found, a release command is sent to all other nodes in the path that are not used for transmission. These released nodes can be reused for exploring additional paths. After receiving the successful acknowledgement, the source node starts to send out sensed data to the successful path with the preassigned path number. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


Figure 3. (a) The found routing path with path circles. (b) The optimized routing path after eliminating path circles.

4.3. Design of two-phase geographic greedy forwarding plus forwarding policies Greedy forwarding mechanism, the main component of geographic routing, usually follows the principle that each node forwards packets to the neighbor node that is closest to the destination with the assumption of highly reliable links. However, this assumption is not realistic. To optimize the forwarding choices, we will observe the performance of TPGFPlus under three different forwarding mechanisms in this paper: (1) Finding an intermediate 1-hop direct neighbor node that is closest to the 2-hop neighbor node. For the first forwarding policy, it will find a neighbor closest to the 2-hop neighbor node and make the maximum progress to the destination, which is commonly employed. But while considered the realistic wireless channel characteristic, this policy may no longer work well. It may choose a neighbor farthest from current node with a poor link [30]. (2) Finding an intermediate 1-hop direct neighbor node that forwards packet from current node to its 2-hop neighbor node with the shortest distance. The second forwarding policy attempts to minimize total geographical distance between the source node and the sink. (3) Finding an intermediate 1-hop direct neighbor node with the most remaining energy, or the best link quality (interference-minimized), or even the optimal multifactor weighted cost function value, and so on. In the third forwarding policy, we adopt Resi_Energy * Distance as a forwarding strategy for our energyaware geographic routing, where Resi_Energy is current energy level of the candidate node and Distance is the distance from current node to the 2-hop neighbor node via the candidate node. In summary, our goal is to study 2-hop geographic node-disjoint multipath routing in duty-cycled WSNs with energy harvesting and find optimal forwarding strategy for different application requirements. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

5. PROPERTIES OF TWO-PHASE GEOGRAPHIC GREEDY FORWARDING PLUS In this section, some basic properties of TPGFPlus algorithm and EC-CKN are discussed, to better understand our cross-layer optimization of the overall system. 5.1. Characteristics of scheduled duty cycle Duty cycle is defined as the ratio between active period and the full active/asleep period. Sensor nodes alternate between dormant and active states [36–39]: In the former, they go to sleep and thus consume little energy; while in the latter, they actively perform sensing tasks and communications, consuming significantly more energy [40]. CKN and EC-CKN are different from other existing duty cycle algorithms for its no waiting delay and simple synchronization mechanism, which we will discuss hereinafter. 1) Avoidance of waiting delay. In a time-varying connectivity network, a message can either be forwarded over the currently awake nodes by using opportunistic routing algorithms or be temporarily buffered in enroute nodes until a better next-hop node wakes up. In the former case, the number of hops may increase significantly, which will incur high energy overheads. In the latter case, the end-to-end latency may increase significantly (e.g., if next-hop node is not scheduled to wake up for many epoches) and the buffering requirements and waking times also increase [9]. Current MAC design for WSNs mostly choose to reduce energy consumption but increases latency, because a sender must wait for the receiver to wake up before it can send out data, which we call ’sleep delay’ caused by the sleep of receiver [41]. However, in our sleep-scheduling layer design, we assume that only once the source node is within the transmission radius of the destination node, it simply waits until the destination wakes up and then hops directly to the destination. But in other cases, we will choose the next-hop node from currently awake

G. Han et al.


Figure 4. Dynamic of duty cycle in energy balanced sleep scheduling scheme.

Figure 5. Two scenarios: (a) scenario I and (b) scenario II.

neighbors. Thus, our design will avoid such waiting delay. For further delay comparison, we will analyze the latency of greedy geographic routing as a function of the number of awake neighbors in the next subsection. 2) Dynamics of duty cycle and simplification of synchronization. On the basis of the present research, if the number of nodes in a network is very small, it may be possible to wake up all nodes for broadcasting through global synchronization with customized active/dormant schedules. For larger-scale WSNs, however, synchronization itself remains an open problem [40]. As shown in Figure 4, to synchronize the sleep schedules of any neighboring nodes, each node wakes up at the beginning of each epoch, which reduces latency and control overhead. Therefore, even for large-scale networks, synchronization in EC-CKN is simple and feasible. 5.2. Theorems and proofs In this subsection, we analyze the relationship between the latency of greedy geographic routing and the number of awake neighbors k, that is, the bounds of expected number of rounds to reach within a specified distance from the

destination in TPGFPlus algorithm under EC-CKN-based network, given in Theorem 1 and 2. Figure 5 shows the two limiting scenarios, in which the destination X is as far away and as close as possible. We assume uniformly random node locations and the disk-r communication model and represent the routing course using a series of discrete progress towards the destination for each step. Notations used in this section: For given k in the EC-CKN algorithm, E.vi / and E.vi /0 are the subsets of N.vi / and N.vi /0 having Erankvj > ˇ Erank ˇ vi , vj 2 N.vi /.ˇ ˇ jN.vi /j, ˇN.vi /0 ˇ, jE.vi /j and ˇE.vi /0 ˇ are the number of the elements in N.vi /, N.vi /0 , E.vi / and E.vi /0 , respectively. ˇ ˇ Let mvi D jE.vi /j C ˇE.vi /0 ˇ be the total number of vi ’s 1-hop and 2-hop neighbor nodes. D is the Euclidean distance from source node S to its destination X and D > r. Ek .D/ is the expected rounds needed to reach the destination. For TPGFPlus algorithm, we divide the range of possible forwarding progress for two hops into t 2 equally-spaced segments. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


Figure 6. Average number of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy one.

Figure 7. Average number of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy two.

Theorem 1. Under TPGFPlus algorithm, the expected rounds to reach within r from the destination is at most 1 D P .m / t1 r 1 i p vi 2mvi t

iD1

(1)

where q2hop is the probability of the neighbor moving farther from the destination q2hop D 2

.mvi /

pi

i

mvi

i t 2r but at most destination

(2)

.m / pi vi

is the probability of the neighbor, among the mvi random 1-hop or 2-hop neighbors of S, moving at least Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

D

Y mvi

iC1 t

2r (i D 0, : : : , t 2) closer to the

1 Y 1 fi A .1 fiC1 / @1 1 fiC1 m 0

(3)

vi

in which fi D

2

cos1 .i = t/

i t

q 1 .i = t/2

D is the Euclidean distance to the destination. And

(4)

G. Han et al.


Proof. According to Theorem 5 in [9], the expected number of rounds to reach within distance r from the destination is at most D 1 P t1 1 r ip q t

iD1

(5)

i

.m /

So what we have to obtain is pi and q, that is, pi vi and q2hop in our Theorem 2. As shown in Figure 5(a), all the 1-hop or 2-hop neighbors of S have the equal probability to move closer or farther from the destination. Thus, q2hop D 2mvi .

Let fi be the probability that a random 1-hop or 2-hop neighbor of S is at least ti 2r closer to the destination X, that is, falls into the slashed area. Through simple computations, we have q Sslashed 2 i 1 2 fi D cos (6) D .i = t/ 1 .i = t/ t 4r2 Using conditional probability formula, we can obtain

.m / pi vi

D

Y mvi

1 Y 1 fi A .1 fiC1 / @1 1 fiC1 m 0

(7)

vi

Figure 8. Average number of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy three.

Figure 9. Average number of paths found by two-phase geographic greedy forwarding (TPGF). Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


.m /

Substituting the parameters pi vi and q2hop into Equation 5, we then obtain the corresponding upper bound. Theorem 2. For the set-up in Theorem 2, the expected rounds to reach within r from the is at least p destination D r

1 t

Pt1

.mvi /

pi

1

.mvi / iD0 .iC1/pi

D

Y mvi

, where q D

0 .1 fiC1 / @1

1 3

C

Y mvi

3 2

mvi

,

1

Proof. By Theorem 6 in [9], the expected number of rounds to reach within distance r from the destination is at most D 1 (10) 1 Pt1 r iD0 .i C 1/pi t

1

1 fi A 1 fiC1

p wi wi 2 cos1 1 C 2 cos1 xi 2 2 (9) p where xi D wi .4 wi /. fi D

(8)

As shown in Figure 5(b) scenario II, the probability of mvi random 1-hop and 2-hop neighbors of S is closer to X than S is

Figure 10. Average hops of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy one.

Figure 11. Average hops of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy two. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


q2hop D

p !mvi 3 1 C 3 2

.m /

(11)

Substituting fi to Equation 7, the corresponding pi vi is obtained. Combined with Equation 10 and Equation 11, the lower bound can be computed.

Similarly, through the geometry calculation, fi can be simplified as follows:

6. PERFORMANCE STUDY

p wi wi 2 cos1 1 C 2 cos1 xi 2 2 (12) p 2 where wi D 1 ti , and xi D wi .4 wi /.

In this section, we present the design details of our simulation. In our previous work, we have proved that TPGF can find more routing paths and have shorter average path length than that of GPSR. So, in this work, we conduct extensive simulations directly compared with TPGF.

1 fi D

Figure 12. Average hops of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy three.

Figure 13. Average hops of paths found by two-phase geographic greedy forwarding (TPGF). Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


Our design has several key features. 1) The impact of MAC layer is ignored in our model. In other words, if the network is collision-free and connected, then each message is delivered. 2) In the simulation experiments, a simple linear energy harvesting model is used. Considering the great variability of environmental energy sources and to their unpredictable nature, [42] has given a comprehensive and detailed energy harvesting model. However, in this paper, our main research focuses on probing

into the impact of harvesting technique on network performance and lifetime enhancement, rather than giving a better design for harvesting model. Therefore, we will simply use a random process to model the energy recharging rate of each node. Nevertheless, it is worth noting that our approach is general and can be applied to any complex and realistic energy harvesting model as well. We assume each node is equipped with a primary battery of capacity Einit . Let Ri be the amount of recharged energy for node vi at each epoch, varying from Rmin to Rmax .

Figure 14. Average number of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy one with adjustable tr .

Figure 15. Average number of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy two with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


Limited to recharging capacity, at any time, the stored energy will never exceed Est . 3) In this work, the energy consumption consists of two parts: the energy consumed by executing ECCKN at each epoch, denoted as EECCKN , and the energy consumed by nodes in the found paths to send and receive packets, denoted as Erouting , that is: Ecost D EECCKN C Erouting

(13)

For EECCKN and Erouting , the energy consumed by nodes to send and receive data is based on the firstorder radio model [43]. And the energy consumed by nodes with the radio in idle mode is approximately the same as the radio in receiving mode [44]. 6.1. Simulation setup We conduct extensive simulation experiments in a new sensor network simulator NetTopo [45], using a network:

Figure 16. Average number of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy three with adjustable tr .

Figure 17. Average hops of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy one with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


800 600 m. The numbers of deployed sensor nodes are increased from 100 to 1000 (each time increased by 100). The value of k is changed from 1 to 10 (each time increased by 1), letting more nodes awake. For every number of deployed sensor nodes, we use 100 different seeds to generate 100 different network deployments. A source node is deployed at the location of (50, 50), and a sink node is deployed at the location of (750, 550). The transmission radius for each node is initially 60 m. Each node is initialized with a certain amount of energy (500 units) before

deployment. The power harvesting rate of each node in one duty cycle is between 0 and 10 units. The energy consumption is set to be 1 unit for executing EC-CKN one time in each node and 5 units for each node in the TPGFPlus routing paths for sending and receiving packets. To evaluate the overall performance of our cross-layer optimized routing protocol, it is necessary to make a comprehensive analysis. The following performance metrics under three different forwarding policies are evaluated:

Figure 18. Average hops of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy two with adjustable tr .

Figure 19. Average hops of paths found by two-phase geographic greedy forwarding plus (TPGFPlus) under the policy three with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm


1) 2) 3) 4)

Average number of found paths. Optimized average hops of found paths. Network sleep rate. Network residual average energy.

G. Han et al.

1) Each node uses a fixed transmission power and is with energy harvesting or without any energy harvesting. 2) Each node can amplify its transmission power to adjust its energy level and is also rechargeable.

6.2. Simulation results As mentioned earlier, we implement our algorithm in different network environments. To illustrate the impacts of our cross-layer optimized framework, we mainly consider two cases:

6.2.1. Fixed transmission power. First, we compare our algorithm TPGFPlus, in which each node is with a fixed transmission power and is rechargeable, with TPGF, in which transmission power is also fixed, but not rechargeable.

Figure 20. Sleep rate of two-phase geographic greedy forwarding plus (TPGFPlus) under the policy one with fixed tr .

Figure 21. Sleep rate of two-phase geographic greedy forwarding plus (TPGFPlus) under the policy one with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


1) The average number of paths of TPGFPlus algorithm under three different forwarding policies for the fixed transmission power case. Figure 6, Figure 7, Figure 8, and Figure 9 show the comparison of explored average number of paths by TPGFPlus under three different policies and TPGF algorithms when the value of k changes for different number of deployed nodes. We can see that TPGFPlus algorithm finds more transmission paths than TPGF algorithm. In addition, TPGFPlus algorithm under the policy three in Figure 8 finds more transmission

paths than other two policies in Figure 6 and Figure 7. Moreover, when networks are sparse, it is hard to find any path because networks are disconnected. On the contrary, when networks are dense and k exceeds a certain value, waking up more sensor nodes cannot always increase the average number of paths. 2) The optimized average hops of paths of TPGFPlus algorithm under three different forwarding policies for the fixed transmission power case. Figure 10, Figure 11, Figure 12, and Figure 13 are the optimized average hops of paths obtained by TPGF-

Figure 22. Sleep rate of two-phase geographic greedy forwarding plus (TPGFPlus) under the policy two with fixed tr .

Figure 23. Sleep rate of two-phase geographic greedy forwarding plus (TPGFPlus) under the policy two with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm


Plus under three different forwarding policies and TPGF algorithms. Both algorithms are not dramatically affected by the changing value of k. However, TPGFPlus algorithm utilizing 2-hop neighborhood information performs better than TPGF algorithm. Furthermore, TPGFPlus algorithm under the policy two in Figure 11 aiming for shortest path performs even better than the other two policies in Figure 10 and Figure 12.

G. Han et al.

3) The comparison of network sleep rate when executing TPGFPlus and TPGF on EC-CKN-based WSNs. Figure 6, Figure 7, Figure 8, Figure 9, Figure 10, Figure 11, Figure 12, and Figure 13 show that TPGFPlus algorithm allows more nodes to sleep, while achieving the same level of average number of paths and average path length compared with TPGF algorithm. we also reveal that the sleep rate of the WSN can strongly affect the transmission performance, for

Figure 24. Sleep rate of two-phase geographic greedy forwarding plus (TPGFPlus) under the policy three with fixed tr .

Figure 25. Sleep rate of two-phase geographic greedy forwarding plus (TPGFPlus) under the policy three with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


example, the routing path can be longer if the WSN has a higher sleep rate. 6.2.2. Adjustable transmission power. Now, we further consider the case where transmission power of each node is adjustable depending on its current energy level. Each node can adjust its power to any discrete value tr (e.g., 60, 70, 80 m), which is also rechargeable. When the current energy of a node is more than the preset value, the node will amplify its transmission power. Thus,

it will consume more energy, balancing the whole network energy consumption, and have more neighbor nodes to choose. In this way, adjustable transmission power affects network connectivity and network sleep rate. 1) The average number of paths of TPGFPlus algorithm under three different forwarding policies for the adjustable transmission power case. Figure 14, Figure 15, and Figure 16 show the explored average number of paths by TPGFPlus under three differ-

Figure 26. The network residual average energy of two-phase geographic greedy forwarding plus (TPGFPlus) under policy one with fixed tr .

Figure 27. The network residual average energy of two-phase geographic greedy forwarding plus (TPGFPlus) under policy two with fixed tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm


ent policies on EC-CKN-based WSN with energy harvesting. Compared with the fixed transmission power case, the adjustable transmission power case performs better, finding more transmission paths in higher density network. 2) The optimized average hops of paths of TPGFPlus algorithm under three different forwarding policies for the adjustable transmission power case. Figure 17, Figure 18, and Figure 19 are the optimized average hops of paths obtained by TPGFPlus under three different forwarding policies. Through compar-

G. Han et al.

ison, we find fixed transmission power case performs better than adjustable one when the network is sparse. But when the network is dense, the adjustable one has better performance, specially under policy three. 3) The network sleep rate when executing TPGFPlus on EC-CKN-based rechargeable WSNs with fixed and adjustable transmission power. Figure 20, Figure 21, Figure 22, Figure 23, Figure 24, and Figure 25 illustrate the network sleep rate for both cases under three different forwarding policies. As we can see, the network sleep rate for adjustable tr case is much

Figure 28. The network residual average energy of two-phase geographic greedy forwarding plus (TPGFPlus) under policy three with fixed tr .

Figure 29. The network residual average energy of two-phase geographic greedy forwarding plus (TPGFPlus) under policy one with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


higher than the fixed tr case. Thus, we can get a more balanced network through enlarging partial nodes’ tr that have much more energy than other nodes. Because those nodes with less energy will have more chance to sleep and thus have enough time to recharge energy. 4) The network residual average energy of TPGFPlus algorithm under three different forwarding policies on EC-CKN-based rechargeable WSNs with fixed and adjustable transmission power. Figure 26,

Figure 27, Figure 28, Figure 29, Figure 30, and Figure 31 show the network residual average energy for fixed transmission power case and adjustable transmission power case of TPGFPlus, respectively. For the fixed tr case, as we can see from the figures, k D 1 does not always prolong network lifetime further. The volatility of curves is largely caused by other uncertainties, such as the dynamic energy harvesting. While executing EC-CKN duty-cycled algorithm in a WSN with energy harvesting, the tra-

Figure 30. The network residual average energy of two-phase geographic greedy forwarding plus (TPGFPlus) under policy two with adjustable tr .

Figure 31. The network residual average energy of two-phase geographic greedy forwarding plus (TPGFPlus) under policy three with adjustable tr . Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


Figure 32. The average multiple curves of Figure 29, Figure 30, and Figure 31.

ditional method that adjusts nodes’ residual energy solely by changing the value of k is not enough. It needs a better control mechanism. Thus, we further introduce the measure of adjustable transmission power. In addition, the network residual average energy under the policy three is more balanced, with less fluctuation. For the adjustable tr case, the residual average energy is slightly lower, because enlarging transmission radio incurs more energy consumption. Also, the network residual average energy for the adjustable tr case is more balanced than the fixed tr case, with less fluctuation. From Figure 32, eliminating the impact of the value of k, we can see the overall trend of network residual average energy is similar. In addition, the network residual average energy under the policy three is more balanced, with less fluctuation.

7. CONCLUSIONS This paper focuses on designing a cross-layer optimized geographic routing that also balances the energy consumption in EC-CKN-applied duty-cycled WSNs with environmental energy harvesting. Its main contributions are the following. First, geographic routing in duty-cycled WSNs should be 2-hop based but not 1-hop based, because of the following: (i) in most existing sleep-scheduling algorithms, it is mandatory for gathering 2-hop neighborhood information and (ii) simulation results in this paper further support this point. Second, cross-layer optimized routing allows more nodes to sleep while achieving the same desired routing performance. Third, we also evaluate the performance of the fixed and adjustable transmission power cases under three different forwarding policies. We make routing decision locally by considering the progress to the destination,

the shortest distance towards the destination, the residual energy level of nodes, and the environmental energy harvesting. In future work, we will give further study on path selection mechanism, for example, selecting paths with minimum inter-interference or other QoS indexes for different applications.

ACKNOWLEDGEMENTS The work is supported by Natural Science Foundation of JiangSu Province of China, no. BK20131137; Science & Technology Pillar Program (Social development) of Changzhou Science and Technology Bureau, no. CE20135052; the Guangdong University of Petrochemical Technology’s Internal Project, no. 2012RC0106; and Jiangsu Province Ordinary University Graduate Innovation Project, no. CXLX13_227. Lei Shu is the corresponding author.

REFERENCES 1. Chen S, Sinha P, Shroff N, Joo C. Finite-horizon energy allocation and routing scheme in rechargeable sensor networks. In Proceedings of INFOCOM’11, Shanghai, China, April 10–15, 2011; 2273–2281. 2. Voigt T, Ritter H, Schiller J. Utilizing solar power in wireless sensor networks. In Proceedings of LCN’03, Bonn, Germany, October 20–24, 2003; 416–422. 3. Miao G, Himayat N, Li YG, Swami A. Cross-layer optimization for energy-efficient wireless communications: a survey. Wireless Communications and Mobile Computing 2009; 9(4): 529–542. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


4. Shu L, Hara T, Nishio S. Multimedia streaming in mobile sensor networks with duty-cycle and environmental energy harvesting. IEEE COMSOC MMTC E-Letter September, 2011; 6(9): 20–24, Available from: http://committees.comsoc.org/mmc/e-news/ELetter-September11.pdf. 5. Zeng K, Ren K, Lou W, Moran PJ. Energy aware efficient geographic routing in lossy wireless sensor networks with environmental energy supply. In QShine’06, Waterloo Ontario, Canada, August 7–9, 2006; 39–51. 6. Ammari HM, Das SK. Joint k-coverage, duty-cycling, and geographic forwarding in wireless sensor networks. In Proceedings of ISCC’09, Sousse, Tunisia, July 5–8, 2009; 487–492. 7. Zorzi M, Member S, Rao R, Member S. Geographic random forwarding (GeRaF) for ad hoc and sensor networks: multi-hop performance. IEEE Transactions on Mobile Computing 2003; 2(4): 337–348. 8. Shu L, Yuan Z, Hara T, Wang L, Zhang Y. Impacts of duty-cycle on TPGF geographical multipath routing in wireless sensor networks. In IWQoS’10, Beijing, China, June 16–18, 2010; 1–2. 9. Nath S, Gibbons P. Communicating via fireflies: geographic routing on duty-cycled sensors. In IPSN’07, Cambridge, Massachusetts, USA, April 25–27, 2007; 440–449. 10. Shu L, Zhang Y, Yang L, Wang Y, Hauswirth M, Xiong N. Tpgf: geographic routing in wireless multimedia sensor networks. Telecommunication Systems 2009; 44(1–2): 79–95. 11. Ruehrup S, Stojmenovic I. Contention-based georouting with guaranteed delivery and minimal communication overhead in wireless ad hoc and sensor networks. In Proceedings of IPDPS, Atlanta, 2010; 1–9. 12. Stojmenovic I, Lin X. Loop-free hybrid singlepath/flooding routing algorithms with guaranteed delivery for wireless networks. IEEE Transactions on Parallel and Distributed Systems October, 2001; 12 (10): 1023–1032. 13. Jurdak R. Signals and communication technology. Wireless Ad Hoc and Sensor Networks 2007: 89–99, Part II. 14. Hoesel L, Nieberg T, Wu J, Havinga P. Prolonging the lifetime of wireless sensor networks by cross-layer interaction. IEEE Wireless Communications Magazine 2004; 11(6): 78–86. 15. Galluccio L, Leonardi A, Morabito G, Palazzo S. A MAC/routing cross-layer approach to geographic forwarding in wireless sensor networks. Ad Hoc Networks August, 2007; 5(6): 872–884. 16. Yuan Z, Wang L, Shu L, Hara T, Qin Z. A balanced energy consumption sleep scheduling algoWirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

rithm in wireless sensor networks. In Proceedings of IWCMC’11, Istanbul, Turkey, July 5–8, 2011; 831–835. 17. Wang H, Wang W, Wu S, Hua K. A survey on the cross-layer design for wireless multimedia sensor networks. Social Informatics and Telecommun Eng, LNCS 2010; 48: 474–486. 18. Stojmenovic I, Nayak A, Kuruvila J, Ovalle-Martinez F, Villanueva-Pena E. Physical layer impact on the design and performance of routing and broadcasting protocols in ad hoc and sensor networks. Computer Communications June, 2005; 28 (10): 1138–1151. 19. Abdu A, Salamah M. Energy-aware transmission scheme for wireless sensor networks. Recent Trends in Wireless and Mobile Networks, Communications in Computer and Information Science 2011; 162: 135–144, Part 1. 20. Tarique M, Tepe KE, Adibi S, Erfani S. Survey of multipath routing protocols for mobile ad hoc networks. Journal of Network and Computer Applications 2009; 32(6): 1125–1143. 21. Caleffi1 M, Paura L. M-DART: multi-path dynamic address routing. Wireless Communications and Mobile Computing 2011; 11(3): 392–409. 22. Shu L, Zhang Y, Yu Z, Yang L, Hauswirth M, Xiong N. Context-aware cross-layer optimized video streaming in wireless multimedia sensor networks. Journal of Supercomputing 2010; 54(1): 94–121. 23. Han G, Choi D, Lim W. Reference node placement and selection algorithm based on trilateration for indoor sensor networks. Wireless Communications and Mobile Computing 2009; 9(8): 1017–1027. 24. Han G, Xu H, Duong T, Jiang J, Hara T. Localization algorithms of wireless sensor networks: a survey. Telecommunication Systems 2013; 52(4): 2419–2436. 25. Han G, Xu H, Jiang J, Shu L, Hara T, Nishio S. Path planning using a mobile anchor node based on trilateration in wireless sensor networks. Wireless Communications and Mobile Computing 2013; 13 (14): 1324–1336. 26. Lee S, Bhattacharjee B, Banerjee S. Efficient geographic routing in multihop wireless networks. In MobiHoc’05, Urbana-Champaign, IL, USA, May 25–28, 2005; 230–241. 27. Wang X, Zhang X, Zhang Q, Chen G. An energyefficient integrated mac and routing protocol for wireless sensor networks. In Proceedings of ICC’09, Dresden, June 14–18, 2009; 1–6. 28. Leong B, Liskov B, Morris R. Geographic routing without planarization. In NSDI, San Jose, CA, May, 2006; 1–25.

G. Han et al.


29. Li C, Hsu W, Krishnamachari B, Helmy A. A local metric for geographic routing with power control in wireless networks. In SECON’05, Santa Clara, California, USA, September 26–29, 2005; 229–239. 30. Seada K, Zuniga M, Helmy A, Krishnamachari B. Energy-efficient forwarding strategies for geographic routing in lossy wireless sensor networks. In Proceedings of SenSys’04, Baltimore, Maryland, November 3–5, 2004; 108–121. 31. Hill J, Szewczyk R, Woo A, Hollar S, Culler D, Pister K. System architecture directions for networked sensors. In Architectural Support for Programming Languages and Operating Systems, New York, USA, 2000; 93–104. 32. Holland M, Wang T, Tavli B, Seyed A, Heinzeilman W. Optimizing physical-layer parameters for wireless sensor networks. ACM Transactions on Sensor Networks February, 2011; 7 (4): 1–20, Article 28. 33. Karp B, Kung H. GPSR: greedy perimeter stateless routing for wireless networks. In Proceedings of the Annual International Conference on Mobile Computing and Networking, Boston, August, 2000; 243–254. 34. Kuhn F, Wattenhofer R, Zhang Y, Zollinger A. Geometric ad-hoc routing: of theory and practice. In Proceedings of the 22nd ACM International Symposium on the Principles of Distributed Computing, New York, USA, 2003; 63–72. 35. Leong B, Mitra S, Liskov B. Path vector face routing: geographic routing with local face information. In Proceedings of the 13th IEEE International Conference on Network Protocols, Washington, USA, 2005; 147–158. 36. Gu Y, Hwang J, He T, Du D. Sense: a unified asymmetric sensing coverage architecture for wireless sensor networks. In IEEE ICDCS, Toronto, June 25–27, 2007; 1–8. 37. Liu J, Zhao F, Cheung P, Guibas L. Apply geometric duality to energy-efficient non-local phenomenon awareness using sensor networks. IEEE Wireless Communications 2004; 11(8): 62–68. 38. Wang X, Xing G, Zhang Y, Lu C, Pless R, Gill C. Integrated coverage and connectivity configuration in wireless sensor networks. In ACM SenSys, New York, USA, 2003; 28–39. 39. Yan T, He T, Stankovic J. Differentiated surveillance service for sensor networks. In ACM SenSys, New York, USA, 2003; 51–62. 40. Wang F, Liu J. Duty-cycle-aware broadcast in wireless sensor networks. In Proceedings of IEEE INFOCOM, Rio de Janeiro, 2009; 468–476. 41. Ye W, Heidemann J, Estrin D. Medium access control with coordinated, adaptive sleeping for wireless sensor networks. IEEE Transactions on Networking 2004; 12(3): 493–506.

42. Dong Y, Shu L, Han G, Guzani M, Hara T. Geographic multipath routing in duty-cycled wireless sensor networks: one-hop or two-hop? In MobiCom’11, Las Vegas, Nevada, USA, September, 2011. 43. Heinzelman W, Chandrakasan A, Balakrishnan H. Energy-efficient communication protocol for wireless microsensor networks. In Proceedings of HICSS’00, USA, January 4–7, 2000; 1–10. 44. Raghunathan V, Schurgers C, Park S, Srivastava M, Shaw B. Energy-aware wireless microsensornetworks. IEEE Signal Processing Magazine 2002; 19(2): 40–50. 45. Shu L, Hauswirth M, Chao H, Chen M, Zhang Y. NetTopo: a framework of simulation and visualization for wireless sensor networks. In Elsevier, Ad Hoc Networks, Netherland, 2010; 799–820.

AUTHORS’ BIOGRAPHIES Guangjie Han is currently an Associate Professor of the Department of Information & Communication System at Hohai University, China. He is also a visiting research scholar of Osaka University from October 2010 to October 2011. He finished the work as a postdoctor of the Department of Computer Science at Chonnam National University, Korea, in February 2008. He received his PhD degree in the Department of Computer Science from Northeastern University, Shenyang, China, in 2004. He has published over 100 papers in related international conferences and journals. He has served as editor of JIT and KSII. He has served as a cochair for more than 20 international conferences/workshops and a TPC member of more than 50 conferences. He has served as a reviewer of more than 50 journals. His current research interests are security and trust management, localization and tracking, and cooperative computing for wireless sensor networks. He is a member of IEEE and ACM. Yuhui Dong is the Patent Examination Cooperation Center of SIPO, Jiangsu, China. She recevied her Master’s degree from the Department of Information & Communication Engineering at the Hohai University, China, in 2012. She received her BS degree in Information and Communication Engineering from Hohai University, China, in 2009. Her current research interests is routing security for Wireless Sensor Networks. Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

G. Han et al.


Hui Guo is currently pursuing her Master’s degree from the Department of Information & Communication Engineering at Hohai University, China. She received her BS degree in Information and Communication Engineering from Hohai University, China, in 2011. Her current research interests are routing and energy analysis for wireless sensor networks.

Lei Shu received the PhD degree in Digital Enterprise Research Institute from National University of Ireland, Galway, Ireland, in 2010. Until March 2012, he was a Specially Assigned Researcher in Department of Multimedia Engineering, Graduate School of Information Science and Technology, Osaka University, Japan. He is a member of IEEE and ACM. Since October 2012, he joined Guangdong University of Petrochemical Technology, China, as a full professor. Since 2013, he started to serve in the following: (i) Dalian University of Technology as a PhD supervisor and (ii) Beijing University of Posts and Telecommunications as a Master Supervisor. Meanwhile, he is also working as the Vice Director of the Guangdong Provincial Key Laboratory of Petrochemical Equipment Fault Diagnosis, China. He is the founder of Industrial Security and Wireless Sensor Networks Lab. His research interests include wireless sensor networks, multimedia communication, middleware, security, and fault diagnosis. He had published over 180 papers in related conferences, journals, and books. He had been awarded the Globecom 2010 and ICC 2013 Best Paper Award. He is serving as Editor-in-Chief for IEEE CommSoft E-letter and EAI Endorsed Transactions on Industrial Networks and Intelligent Systems and Associate Editors for the following: (i) Elsevier, Journal of Network and Computer Applications; (ii) Wiley, Transactions on Emerging Telecommunications Technologies; (iii) Wiley, Wireless Communications and Mobile Computing; (iv) Wiley, Security and Communication Networks; (v) Wiley, International Journal of Communication Systems; (vi) IET Communications; (vii) IET Networks; (viii) IET Wireless Sensor Systems; (ix) KSII Transactions on Internet and Information Systems (TIIS); (x) Inderscience, International Journal of Sensor Networks; (xi) Ad Hoc & Sensor Wireless Networks; (xii) Journal of Internet Technology; (xiii) Journal of Communications; (xiv) Guest Editor: IEEE Wireless Communication Magazine; and (xv) Guest Editor: IEEE System Journal. He is serving as the Vice Chair for SIG on Energy Harvesting Communications in IEEE Technical Committee on Green Communications & Computing and serving as a member in IEEE MMTC Service and Publicity Board. He served

Wirel. Commun. Mob. Comput. (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/wcm

to more than 50 various Cochair for international conferences/workshops, for example, IWCMC, ICC, ISCC, ICNC, and Chinacom, especially Symposium Cochair for IWCMC 2012, ICC 2013, General Chair for Chinacom 2014, Steering Chair for InisCom 2015, TPC member of more than 150 conferences, for example, DCOSS, MASS, ICC, Globecom, ICCCN, WCNC, and ISCC. Dapeng Wu (S’98-M’0-4SM’06F’13) received BE in Electrical Engineering from Huazhong University of Science and Technology, Wuhan, China, in 1990; ME in Electrical Engineering from Beijing University of Posts and Telecommunications, Beijing, China, in 1997; and PhD in Electrical and Computer Engineering from Carnegie Mellon University, Pittsburgh, PA, USA, in 2003. He is a professor at the Department of Electrical and Computer Engineering, University of Florida, Gainesville, FL, USA. His research interests are in the areas of networking, communications, signal processing, computer vision, and machine learning. He received the University of Florida Research Foundation Professorship Award in 2009, AFOSR Young Investigator Program (YIP) Award in 2009, ONR YIP Award in 2008, NSF CAREER award in 2007, the IEEE Circuits and Systems for Video Technology Transactions Best Paper Award for Year 2001, and the Best Paper Awards in IEEE GLOBECOM 2011 and International Conference on Quality of Service in Heterogeneous Wired/Wireless Networks (QShine) 2006. Currently, he serves as an Associate Editor for IEEE Transactions on Circuits and Systems for Video Technology, Journal of Visual Communication and Image Representation, and International Journal of Ad Hoc and Ubiquitous Computing. He is the founder of IEEE Transactions on Network Science and Engineering. He was the founding Editor-in-Chief of Journal of Advances in Multimedia between 2006 and 2008 and an Associate Editor for IEEE Transactions on Wireless Communications and IEEE Transactions on Vehicular Technology between 2004 and 2007. He is also a Guest Editor for IEEE Journal on Selected Areas in Communications, Special Issue on Cross-layer Optimized Wireless Multimedia Communications. He has served as Technical Program Committee (TPC) Chair for IEEE INFOCOM 2012 and TPC chair for IEEE International Conference on Communications (ICC 2008), Signal Processing for Communications Symposium, and as a member of executive committee and/or technical program committee of over 80 conferences. He has served as Chair for the Award Committee and Chair of Mobile and wireless multimedia Interest Group (MobIG), Technical Committee on Multimedia Communications, and IEEE Communications Society. He was a member of Multimedia Signal Processing Technical Committee, IEEE Signal Processing Society from 1 January 2009 to 31 December 2012. He is an IEEE Fellow.