Anda di halaman 1dari 9

www.ietdl.

org
Published in IET Wireless Sensor Systems Received on 19th October 2010 Revised on 20th January 2011 doi: 10.1049/iet-wss.2010.0084

ISSN 2043-6386

Mobility-based clustering protocol for wireless sensor networks with mobile nodes
S. Deng1,2 J. Li1 L. Shen1
1

National Mobile Communications Research Laboratory, Southeast University, Nanjing, Jiangsu 210096, Peoples Republic of China 2 Department of Physics and Telecommunications Engineering, Hunan City University, Yiyang, Hunan 413000, Peoples Republic of China E-mail: lfshen@seu.edu.cn

Abstract: In this study, the authors propose a mobility-based clustering (MBC) protocol for wireless sensor networks with mobile nodes. In the proposed clustering protocol, a sensor node elects itself as a cluster-head based on its residual energy and mobility. A non-cluster-head node aims at its link stability with a cluster head during clustering according to the estimated connection time. Each non-cluster-head node is allocated a timeslot for data transmission in ascending order in a time division multiple address (TDMA) schedule based on the estimated connection time. In the steady-state phase, a sensor node transmits its sensed data in its timeslot and broadcasts a joint request message to join in a new cluster and avoid more packet loss when it has lost or is going to lose its connection with its cluster head. Simulation results show that the MBC protocol can reduce the packet loss by 25% compared with the cluster-based routing (CBR) protocol and 50% compared with the low-energy adaptive clustering hierarchy-mobile (LEACH-mobile) protocol. Moreover, it outperforms both the CBR protocol and the LEACH-mobile protocol in terms of average energy consumption and average control overhead, and can better adapt to a highly mobile environment.

Introduction

Wireless sensor networks (WSNs) consist of a large number of sensor nodes that are densely deployed in a region of interest to collect data about a target or event, and provide a variety of sensing and monitoring applications [1]. For many applications, such as sea exploration, wildlife protection, and trafc congestion control, mobile sensor nodes need to be deployed in a network, which would cause frequent topology change and thus result in high packet loss. For this reason, mobility becomes a critical issue that must be considered in the design of WSNs with mobile nodes [2, 3]. Sensor nodes are constrained in energy, which has posed many challenges on the design of a sensor network [4, 5]. Clustering is a technique that can effectively reduce the energy consumption of sensor nodes and has been widely used in WSNs [6, 7]. A variety of clustering protocols have been proposed to address the energy efciency problem in different network scenarios [8 10]. However, most existing clustering protocols are unable to support mobile sensor nodes because they do not consider the nodes movements after clustering. The low-energy adaptive clustering hierarchy-mobile (LEACH-mobile) protocol proposed in [11] supports the mobility of a sensor node by adding its membership declaration to the LEACH protocol. The idea behind this membership declaration is to conrm the inclusion of sensor nodes in a particular cluster in the steady-state phase. The sensor node, which does not receive a request from its
IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 3947 doi: 10.1049/iet-wss.2010.0084

cluster head during two consecutive frames, can also recognise that it has moved out of the cluster, and thus it will broadcast a cluster joint request message in order to join in a new cluster and avoid losing more packets. Therefore the LEACH-mobile protocol increases the successful packet delivery rate at the cost of increased control overhead. Santhosh et al. [12] proposed a LEACH-mobile-enhanced protocol based on a mobility metric remoteness for clusterhead election with the argument that it is better to elect a node as a cluster head that has less mobility relative to its neighbours. The node with the minimum mobility or in group motion with other members is more likely to be elected as a cluster head. Lan et al. [13] argued that the remaining energy of a cluster-head node and the total number of nodes in a cluster, together with the distance between a cluster-head node and a non-cluster-head node should be taken into account during clustering. Samer et al. [14] proposed an adaptive TDMA scheduling and round-free cluster-head protocol called cluster-based routing (CBR) protocol. In the CBR protocol, a cluster-head receives data not only from its member during the allocated TDMA timeslot, but also from other sensor nodes that just enter the cluster when it has a free timeslot. The CBR protocol changes TDMA scheduling adaptively according to the dynamic trafc and mobility conditions in the network. As a result, the CBR protocol can signicantly reduce the packet loss compared to the LEACH-mobile protocol. However, the control overhead used in the CBR protocol is higher than that in the
39

& The Institution of Engineering and Technology 2011

www.ietdl.org
LEACH-mobile protocol because of the control messages used. Both the LEACH-mobile and CBR protocols aim at increasing the successful packet delivery rate at the cost of increased control overhead, which may cause more energy consumption and reduce the network lifetime. Santhosh et al. [12] and Lan et al. [13] focus on the stability of each cluster by considering the mobility of each node and the distance between a non-cluster-head sensor node and a cluster-head node. However, this may result in unfavourable cluster-head selection because the two sensor nodes may have opposite moving directions. To address this problem, we propose a mobility-based clustering (MBC) protocol for WSNs with mobile nodes in this paper. The proposed protocol will take an estimated connection time into account in order to build a more reliable path depending on the stability or availability of each link between a non-cluster-head sensor node and a clusterhead node. In the MBC protocol, a node elects itself as a cluster head based not only on its residual energy but also on its mobility in order to achieve balanced energy consumption among all nodes and thus longer lifetime of the network. During clustering, a non-cluster-head node takes into account its connection time with and the distance from a cluster head, and the residual energy and node degree of the cluster head, which can guarantee a stable link with a cluster head and thus increase the successful packet delivery rate, and reduce the control overhead and energy consumption because of the less frequent membership changes. A cluster head creates a TDMA schedule based on the estimated connection time of each node in its cluster in an ascending order. The non-cluster-head nodes send data packets according to the time schedule. When a sensor node has lost or is going to lose connection with its cluster head, it will broadcast a joint request message in order to join a new cluster and avoid more packet loss. The rest of the paper is organised as follows. The network model, together with mobility model and radio model are described, and the MBC protocol is presented in Section 2. In Section 3, the performance of the proposed protocol is evaluated through simulation results. Finally, conclusions are presented in Section 4.

MBC protocol

In this section, we present an MBC protocol for a WSN with mobile nodes. We will rst introduce the network model under study, and the mobility model and radio model used in the study, and then we will describe the characteristics and procedures of the protocol. 2.1 Network model

We consider a sensor network shown in Fig. 1, where the sensor nodes are clustered into a group of clusters, and periodically collect and transmit sensed data about a target or an event in a region of interest to a remote observer (for example, a base station). Each sensor node within a cluster transmits its data to its cluster head within one transmission hop. The cluster head performs data aggregation on the data it receives from its cluster members and then transmits the aggregated data to the base station along a multi-hop path. Note that this paper focuses on clustering. Multi-hop routing is beyond the scope of this paper. The network can be modelled as an undirected connectivity graph G (V, E ), where V is a nite set of nodes, and (i, j ) [ E represents a wireless link between node i and node j. A function, which takes into account the position, speed, moving direction and transmission range, is used to indicate a sensor nodes condition in the network in a Cartesian coordinate, that is Fi (t ) = f ((x(i, t ), y(i, t )), v(i, t ), u(i, t ), Ri ) (1)

where (x (i, t ), y(i, t )) is the position, v (i, t ) is the speed, u(i, t )

Fig. 1 Sensor information forwarding with clustering and aggregation


40 IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 39 47 doi: 10.1049/iet-wss.2010.0084

& The Institution of Engineering and Technology 2011

www.ietdl.org
is the moving direction of node i at time t and Ri is the communication range of node i. If node j is a neighbour of node i, the relative function can be expressed as Fj i (t ) = g ((x(j i, t ), y(j i, t )), v(j i, t ), u(j i, t ), Ri , Rj ) (2) where (x ( j_i, t ), y( j_i, t )) is the relative position, v ( j_i, t ) is the relative speed, u( j_i, t ) is the relative moving direction of node j to node i at time t and Ri , Rj are the communication range of node i, j, respectively. 2.2 Random waypoint model transmitter and a receiver can be calculated as Etran (k , d ) = Eelec k + eamp k d a Erece (k ) = Eelec k (3) (4)

There exist many kinds of mobility model that attempt to realistically represent the behaviours of mobile nodes. The random waypoint model in [14] is a commonly used synthetic model for mobility, for example, mobile ad hoc and sensor networks. It is an elementary model, which describes the movement pattern of independent nodes by simple terms. In our discussion, we use the random way point model as the mobility model and the parameters are set as below [11, 14]: Q: two-dimensional (2D) deployment region; finit (x): the initial node spatial distribution; ps (0 ps 1): the percentage of static nodes; tp: the pause time of each node; vmin: the minimum speed of each node; vmax: the maximum speed of each node; Direction of each node between [0, 2p).

where Etran and Erece are the transmitting cost and receiving cost, respectively, Eelec is the energy dissipation to run the transmitter or receiver circuitry, and eampd a is the energy dissipation for the transmission amplier. The propagation exponent a can be from 2 to 4 depending on the transmission condition between the transmitter and receiver [15]. For the experiment described in this paper, the communication energy parameters are set as: Eelec 50 nJ/bit, eamp 10 pJ/bit/ma. In addition, we set a 2.5 because of the maximum transmission range and our assumed channel condition in the experiment. Moreover, a threshold Pr-thresh is specied for the received power of a data packet as in [7]. If the received power is below the threshold, the packet is discarded because it cannot be detected and received successfully. Otherwise, the packet can be received successfully. In addition, we assume that the radio channel is symmetric, that is, the energy required to transmit a message from node i to node j is the same as that from node j to node i for a given SNR. 2.4 Assumptions

For the proposed clustering protocol, we make the following assumptions: 1. All sensors in the network are homogenous. The sensors have the same physical characteristics, the same energy capacity and the same transmission range at the time of network deployment. 2. Each sensor in the network knows its own location and velocity. 3. The base station is stationary. 4. All sensors in the network are time synchronised. 5. Each sensor node can estimate the time it takes to transmit a packet, which is dened as the ratio of the packet length and the transmission rate. Considered the constraint in energy supply and bandwidth in WSN, some assumptions such as location, velocity estimation and time synchronisation mentioned above can be achieved by using some location and synchronisation methods in [16 18]. 2.5 Protocol description

In this model, each mobile node selects a random waypoint in the 2D deployment region Q, and then it moves towards that waypoint from current waypoint with speed v between [vmin , vmax]. If the node hits the boundary, it reects off the boundary with the same speed. The reection involves a change in direction of the incident ray. Let the angle of incidence be the angle between a normal line drawn to the boundary and the incident ray, and the angle of reection be the angle between this normal line and the reected ray, and then the angle of incidence equals to the angle of reection. Node waits a random pause time, tp , at each waypoint before starting to move towards the next waypoint. The process is shown in Fig. 2. 2.3 Radio energy dissipation model

Without considering the shadowing losses in our simulation, we use the same radio energy dissipation model in [7] as the radio model. In this model, the costs for transmitting and receiving a k-bit message through a distance d between a

Fig. 2 Random waypoint model


IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 3947 doi: 10.1049/iet-wss.2010.0084

Similar to the LEACH-mobile protocol [11] and the CBR protocol [14], our proposed MBC protocol can also be divided into two phases: the set-up phase when a set of clusters are organised and a TDMA schedule is created, followed by a steady-state phase when data are delivered to the base station. In the LEACH-mobile protocol and the CBR protocol, a non-cluster-head sensor node chooses a cluster head with the minimum amount of transmission energy required for communication between each other. A packet transmitted by a sensor node is considered lost when the sensor node moves out of its cluster. Thus, if the sensor node changes its membership frequently (for example, because of the high mobility), packet loss would be signicant. Moreover,
41

& The Institution of Engineering and Technology 2011

www.ietdl.org
the joint request messages sent to avoid more packet loss also lead to more energy consumption and control overhead. To address such problems, the reliability of a path depending on the stability or availability of each link is taken into account in our proposal. Fig. 3 gives the slot structure of the proposed protocol. In the set-up phase, a non-cluster-head node searches for a stable link to a cluster head in order to increase the successful packet delivery rate. In the steady-state phase, a sensor node broadcasts a joint request message in order to reduce the packet loss resulted from the disconnection between a cluster head and a non-cluster-head node when the connection time is over. It chooses a new cluster based on a value related to each cluster head. The value is used to indicate its suitability to connect with such cluster head. Moreover, since both the cluster-head and the non-clusterhead sensor nodes maintain the information on the estimated connection time, at the end of the transmission, both of them will check whether the sensor node is going to remain in the cluster when its next timeslot comes. The cluster head will eliminate the sensor node from the TDMA schedule whereas the sensor node will broadcast a joint request message if the node is not going to remain in the cluster. After a new member is accepted, the cluster head will adjust the TDMA schedule and broadcast the schedule to the members within its cluster. 2.5.1 Set-up phase: In this phase, a sensor node elects itself as a cluster-head if its generated random value between 0 and 1 is smaller than a threshold T (n)new . The threshold can be comprehended as the probability to elect a cluster head, and it is determined based on a generalised formula that takes into account not only the nodes residual energy but also its mobility. Each sensor node in the network will rst choose the most suitable cluster head to join based on the information it receives from the cluster heads. Once a cluster head is chosen, the sensor node will send a registration message to inform that cluster head. After the cluster head receives the registration message from a node that would like to join the cluster, it will create a TDMA schedule and assign the node a timeslot for transmission. 1. Cluster-head election: In LEACH protocol presented in [7], the threshold is formulated as follows T (n) = p , 1 p [r mod(1/p)] n [ G (5) heads in the most recent rmod(1/p) rounds. Using this threshold, each node will be a cluster head at some point within 1/p rounds. During round 1 (r 0), each node has a probability p of becoming a cluster head. The nodes that are cluster heads in round 1 cannot be cluster heads for the next 1/p 2 1 rounds. After 1/p 2 1 rounds, T (n) 1 for any nodes that have not yet been cluster heads, and in 1/p rounds, all nodes are once again eligible to become cluster heads. Some protocols [6, 7] consider the residual energy of a sensor node available at the time of cluster-head election in order to increase the lifetime of a network. The node with the largest residual energy is selected as a cluster head. Other protocols [12, 13] consider a nodes mobility in cluster-head election. The node with the smallest relative mobility is selected as a cluster head. In our proposed protocol, we consider both the residual energy and the current speed of each sensor node in clusterhead election in order to avoid that low-energy nodes are selected as cluster heads and balance the energy consumption among all nodes. Thus, the threshold T (n) is multiplied by the factors representing the residual energy and the current speed of a node, that is T (n)new = p 1 p [r mod(1/p)] En current vmax vn current , Emax vmax

n [ G

(6)

where p is the rate between the number of cluster heads and the number of total sensor nodes, 1/p is the expected number of nodes in one cluster, r is the index of the current round and G is the set of nodes that have not been cluster

where En_current is the current energy, vn_current is the current speed, Emax is the initial energy and vmax is the maximum speed of the node. Using this threshold, the nodes with more residual energy and lower speed may have more probability to be elected as cluster heads. Moreover, we use (vmax 2 vn_current)/vmax instead of vmax/vn_current to avoid T (n) growing larger than 1. The more residual energy makes sure that the cluster heads are capable of prolonged extra work and the lower speed cope with the situation of cluster head going out of reach because of high speed. In some mobile environment with LEACH-mobile and CBR, cluster heads are assumed to be stationary and static through different rounds. In this paper, we consider a more general scenario and assume that cluster heads may change their positions in different rounds. 2. Advertisement: After a cluster head is selected, it broadcasts an advertisement message as well as its location, velocity to the sensor nodes within its transmission range using a carrier sense multiple access with collision avoidance (CSMA/CA) media access control (MAC) protocol. Since all cluster heads have no idea about their neighbours, they use the maximum transmission energy when broadcasting the advertisement messages in order to inform as many sensor

Fig. 3 MBC slot structure


42 IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 39 47 doi: 10.1049/iet-wss.2010.0084

& The Institution of Engineering and Technology 2011

www.ietdl.org
nodes as possible. In this phase, non-cluster-head sensor nodes must keep their receivers ON in order to receive the messages from the cluster heads. 3. Decision: On receiving the advertisement messages from one or more cluster heads, a non-cluster-head sensor node will decide which cluster it would like to join. Like LEACH, in both LEACH-mobile and CBR, a sensor node chooses a cluster head with the minimum amount of transmission energy required for communication between each other based on the signal strength of the received advertisement messages. To conrm the successful connection between the sensor node and the cluster head during the steady-state phase, LEACH-mobile and CBR use a simple and traditional method that adds membership declaration of mobile nodes to the original LEACH protocol. A packet is considered lost when a non-clusterhead sensor node moves out of its cluster. However, both LEACH-mobile and CBR still have high packet loss because of frequent topology changes, especially in a highly mobile environment. Additionally, such changes will lead to more control overhead and energy consumption because a sensor node needs to broadcast a request to join a new cluster to avoid more packet loss. The main reason for this problem is that the sensor node may join a cluster of which it will move out in a short time. To signicantly reduce the packet loss and energy consumption, the reliability of a path depending on the stability or availability of each link should be taken into account. In the MBC protocol, each member node is assigned a value related to a cluster head that is used to indicate its suitability to connect such cluster head. This value is determined based on a generalised formula that takes into account the estimated connection time, together with the residual energy and node degree of the cluster head, and the distance between the sensor node and the cluster head. We assume that the Cartesian position of node i at time t is i: (xi + vi cos ui t , yi + vi sin ui t ) and cluster head j is Thus, the value can be formulated as j: (xj + vj cos uj t , yj + vj sin uj t ) (8) Wij = a Ej current dij Dtij +b 1 +c Emax Nj current Rtran tframe (12) where Wij is the value assigned to sensor node i, which indicates its suitability for connection with cluster head j, Ej_current is the current energy of cluster head j, Nj_current is the number of members of cluster j after node i joins in, dij is the distance between sensor node i and cluster head j, Dtij is the estimated connection time between node i and j, tframe is the expected duration of a data frame. The factor Ej_current/EmaxNj_current denotes the average normalised energy that cluster head j can provide to each member. The rate 1 2 dij/Rtran is used instead of Rtran/dij in order to avoid the factor growing larger than 1. If Nexp is used to represent the expected number of members in one cluster, that is Nexp = Ntotal p Ntotal 1 p = p Ntotal p (13) (7) also assume that in each frame the node occupies a given slot and n slots is necessary for data transmission for any node i. Since the node will disconnect the connection when it nishes transmission, we can limit t [ [0, nt(1/p 2 1)], where t is the duration of each timeslot, nt(1/p 2 1) denotes the maximum connection time of each node. Further analysing inequality (10), we can obtain the following results: 1. If node i and node j have the same velocity, that is, v (i, j ) 0, u(i, j ) 0, inequality (10) turns to (9), and thus each t [ [0, nt (1/p 2 1)] satises inequality (10). 2. If node i and node j have different velocities, that is, v (i, j ) = 0, u(i, j ) = 0, let t1 and t2 represent the two roots of the function (xj + vj cos uj t xi vi cos ui t )2 + (yj + vj sin uj t yi vi sin ui t )2 = R2 tran Then, we have Case 1: max{t1, t2} 0, no t satises inequation (10); Case 2: 0 , max{t1 , t2}, nt (1/p 2 1), we have t [ [0, max{t1 , t2}]; Case 3: max{t1 , t2} nt (1/p 2 1), we have t [ [0, nt (1/p 2 1)]. Let Dtij represent the length of interval satisfying (10). Obviously, we have Dtij 0 for Case 1. Thus, Dtij denotes the connection time between node i and node j. The dimensionless value, which is used to indicate suitability of member to connect with a cluster head, is generalised as a linear combination with constant coefcients a, b and c between 0 and 1. The coefcients a, b and c represent the signicance of each factor and a+b+c=1 (11)

where (x, y) is the initial Cartesian position, v is the speed and u is the moving direction. The subscript i, j correspond to node i and node j, respectively. For the ease of exposition, we use the same symbol to denote a cluster and its cluster head. We set the time that a sensor node receives an advertisement message from one cluster head as t 0, and thus we have (xj xi )2 + (yj yi )2 R2 tran (9)

where Rtran represents the maximum communication range. If node i remains in cluster j at time t, we have (xj + vj cos uj t xi vi cos ui t )2 + (yj + vj sin uj t yi vi sin ui t )2 R2 tran (10)

The expected number of nodes in one cluster is 1/p, and thus each data frame can be divided into 1/p 2 1 timeslots, and each timeslot is allocated to one member. We assume that each node always has data to transmit during the steadystate phase, and the amounts of data are also the same. We
IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 3947 doi: 10.1049/iet-wss.2010.0084

where Ntotal is the number of sensor nodes in the network,


43

& The Institution of Engineering and Technology 2011

www.ietdl.org
Lpacket is the length of a data packet, tframe can be expressed as tframe = Lpacket Lpacket 1 p Nexp = Rb Rb p (14) data packet, the packet is also considered lost if it is not completely transmitted, which may result from a sudden change in the speed or direction of a sensor node, network congestion or hardware failure during the transmission, and therefore acknowledgments (ACKs) are necessary during the transmission. After receiving a data packet successfully, the cluster head will send an ACK message to the non-cluster-head sensor node. The sensor node will conrm that the packet has been successfully transmitted if it receives the ACK message, or it will broadcast a cluster joint request message. The cluster head upon receiving a cluster joint request message will transmit a cluster-head advertisement message like in the set-up phase to that node. Thus, to further reduce the control overhead, we can omit membership declaration used in LEACH-mobile and CBR. Instead, a cluster head will wait to receive sensed data according to the TDMA schedule in each timeslot. In addition, since both the cluster-head and the non-clusterhead sensor node maintain the information on the estimated connection time, both of them can check whether a sensor node is going to remain in the cluster when its next timeslot comes at the end of the transmission. If the sensor node is not going to remain in the cluster, it will broadcast a request message to join a new cluster and avoid more packet loss before it loses connection with the cluster head. On the other hand, the cluster head will remove the sensor nodes membership. Fig. 4a illustrates the process of a node that leaves an old cluster and joins a new one. Node 6 joins cluster J, whereas node 9 leaves the cluster. The cluster head of cluster J removes node 9 from and adds node 6 into the TDMA schedule. The TDMA schedule is adjusted based on the estimated connection time Dt between a non-cluster-head node and a cluster-head node in an ascending order. The new TDMA schedule may has a form as shown in Fig. 4b. The timeslot substitution of node 9 by node 6 will lead to a higher successful packet delivery rate and a better channel utilisation. The whole protocol can be described by Algorithms in Figures of Appendix.

where Rb is the transmission bit rate. In (12), the rst factor mainly pays attention on the fairness during clustering, coping with the situation that a cluster head with less residual energy may have a larger number of members. The second one focuses on a short distance between cluster head and member node, and the third one ensures that member node has longer connection time with its cluster head. Each sensor node will choose the cluster head with the largest value to join according to (12). After the node decides the cluster it will join and send a registration message to inform the cluster head. This registration message is sent to the cluster head by using a CSMA/CA MAC protocol. In this phase, all cluster heads must keep their receivers on. 4. Schedule creation: After a cluster head receives the registration messages from the nodes that would like to join the cluster, it will create a TDMA schedule based on the number of nodes and assign each node a timeslot for data transmission. We assume that n data frames are transmitted consecutively during the steady-state phase. Node i begins its transmission at (ni + k(1/p 2 1))t (0 k n), where ni denotes the sequence number of node i in the transmission schedule, t is a constant that denotes the duration of a timeslot. Thus, with the constraint (ni + k(1/p 2 1))t Dtij , the created TDMA schedule is an ascending sort of Dtij of each sensor node to ensure that each sensor node is able to send data packets successfully in one cluster as many as possible, that is, maximise k. 2.5.2 Steady-state phase: In our proposed protocol, we assume that all nodes are time synchronised and start the set-up phase at the same time. A non-cluster-head node always has data to send to the cluster head in its allocated timeslot. Since the energy dissipation caused by wireless communication modules is signicantly larger than that caused by sense modules, sleep/wakeup scheduling is widely used in WSNs and many algorithms have been proposed to cope with the status switch between sleep and active [19]. By using a heuristic mechanism mentioned, each sensor node wakes itself up one timeslot before its scheduled timeslot according to the TDMA schedule and goes back to a sleep mode after its timeslot. Either cluster head or member node can adjust its transmission power according to the distance from its neighbour. The distance can be calculated by using the nodes position information contained in the received message. In contrast, the cluster head must keep its radio on to send data request messages, receive data from the sensor nodes, and send and receive other messages needed to maintain the network. In both LEACH-mobile and CBR, a data packet is considered lost when a cluster head does not receive data from a member node. In this case, the cluster head will remove the member node from its transmission schedule. On the other hand, if the member node does not receive a data request from its cluster head, it can realise that it has moved out of its cluster and will broadcast a cluster joint request message. Because of the transmission duration of a
44

Fig. 4 New adaptive TDMA scheduling creation for MBC


a Mobile nodes join a new cluster b Old and new adaptive TDMA scheduling IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 39 47 doi: 10.1049/iet-wss.2010.0084

& The Institution of Engineering and Technology 2011

www.ietdl.org
3 Performance evaluation
In this section, we evaluate the performance of our proposed clustering protocol through simulation results. For evaluation, we compare the MBC with the LEACH-mobile protocol and the CBR protocol in terms of the percentage of data packets that are successful in reaching cluster heads from each node and the average energy consumption for successfully receiving a data packet. We formulate successful packet delivery rate Rreceived , average energy consumption Eavg and average control overhead Oavg as below Rreceived =
def

Nreceived Nreceived + Nlost Etotal Nreceived

(15)
Fig. 5 Average energy consumption against the percentage of cluster heads

Eavg = Oavg =
def

def

(16)

Ototal N Lctrl = ctrl Nreceived Nreceived

(17)

In (15) (17), Nreceived is the number of successfully received packet, Nlost is the number of lost packet, Ptotal is the total energy consumption in the network, Ototal is the total control overheads in the network, Nctrl is the number of control packets and Lctrl is the length of a control packet. Table 1 gives the parameters used in the simulation experiments. According to [7], we set threshold Pr-thresh 0.0013 J/bit because of our assumed channel condition in the experiment. The mentioned transmission range represents the distance used in transmission between cluster head and member node; however, the transmission range among cluster heads or between cluster head and base station can be larger. The nodes use the transmission power matched the minimum distance if the distance between cluster head and member node is smaller than 20 m, whereas they can adjust the transmission power matched the current distance if the distance is between 20 and 80 m. In our simulation, the size of network is 2000 2000 m2 and the transmission range among nodes is 20 80 m within cluster, and thus 625 4000 nodes are needed for coverage for 2D uniform deployment. Considering the complex and real application, we select 1500 nodes in the simulation. Fig. 5 shows that the average energy consumption, which mainly results from the transmission between cluster head and member node, varies with the increase of the rate between the number of cluster heads and the number of total sensor nodes, using parameters given in Table 1. It is seen that the energy consumption decreases as the rate increases; however, the curve attens when the rate is larger than 5%. Moreover, the distance between cluster heads is around 160 m (maximum transmission range equals to
Table 1
Parameter network size number of sensor nodes node deployment mode length of a data packet length of a control packet bit rate transmission range within cluster Simulation parameters Value 2000 2000 m2 1500 2D uniform distribution 2000 bits 100 bits 9600 bps 2080 m

80 m) on the assumption that sensor nodes are uniformly deployed, and thus the rate should be larger than 3% in order to satisfy network coverage. Therefore we use 5% in latter simulation, namely, p 0.05. In Figs. 6 8, we vary the number of mobile sensor nodes with the maximum speed of a node at 25 m/s, whereas in

Fig. 6 Percentage of received packets against the number of mobile nodes

Fig. 7 Average energy consumption per packet against the number of mobile nodes
45

IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 3947 doi: 10.1049/iet-wss.2010.0084

& The Institution of Engineering and Technology 2011

www.ietdl.org

Fig. 8 Average overhead per packet against the number of mobile nodes

Fig. 10 Average energy consumption per packet against speed

Figs. 9 11 we vary the speed of the nodes with 100% mobile sensor nodes. Fig. 6 shows the percentage of successfully received packets with the MBC protocol, LEACH-mobile and CBR, respectively. It is seen that the MBC protocol can achieve a better successful packet delivery rate than both LEACHmobile and CBR because of the stable link between a sensor node and its cluster head and the ascending order in the TDMA schedule. Although the successful packet delivery rate decreases as the number of mobile nodes increases for all three protocols because of the increase of broken links, the rate with the MBC protocol decreases slower than that with either LEACH-mobile or CBR. Fig. 7 shows the average energy consumption for successfully receiving a packet with the MBC protocol, LEACH-mobile and CBR, respectively. In the MBC protocol, a sensor node chooses a cluster with the longest connection time rather than one with the minimum distance, which leads to a little more energy consumption because the sensor node may not choose the nearest cluster head. However, this can signicantly reduce the amount of request messages by omitting membership declaration during the steady-state phase. In addition, a more stable link also leads to fewer membership changes and less control

Fig. 11 Average overhead per packet against speed

Fig. 9 Percentage of received packets against speed


46

overhead. Therefore the MBC protocol consumes less power compared to both LEACH-mobile and CBR. Fig. 8 shows the control overhead for successfully receiving a data packet with the MBC protocol, LEACHmobile, and CBR, respectively. Although the CBR protocol uses more control messages compared to the LEACHmobile protocol, the total control overheads with the two protocols remain close because of the increased successful transmission with the CBR protocol. Meanwhile, the MBC protocol uses fewer control messages because of two reasons: the joint request messages sent to avoid more packet loss are reduced because a sensor node changes its membership less frequently, and the omission of membership declaration also reduces the overhead. In addition, in all three protocols, the control overhead increases as the number of mobile sensors increases. Figs. 10 and 11 show that the MBC protocol outperforms both the LEACH-mobile protocol and the CBR protocol in terms of successful packet delivery rate, average energy consumption and average control overhead as the speed of a sensor node increases, indicating that the MBC protocol can better adapt to a highly mobile environment. In Figs. 10 and 11, it is seen that big chunks of improvement in terms of energy consumption and control overhead exist in case of almost stationary nodes. The former one results from two aspects: (i) decrease in the amount of request messages by
IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 39 47 doi: 10.1049/iet-wss.2010.0084

& The Institution of Engineering and Technology 2011

www.ietdl.org
omitting membership declaration during the steady-state phase, (ii) power adaptation during transmission. On the other hand, the latter one, namely, the decrease in control overhead in case of almost stationary nodes is mainly because of the omission of membership declaration.
13 Computing and Communications, San Francisco, American, December 2008, pp. 248 253 Lan, T.N., Xavier, D., Razvan, B., et al.: An energy efcient routing scheme for mobile wireless sensor networks. Proc. IEEE Int. Symp. on Wireless Communication Systems, Reykjavik, Iceland, October 2008, pp. 568 572 Samer, A.B., CheeK, N.K., Mohd, F.R., et al.: Cluster based routing protocol for mobile nodes in wireless sensor network. Proc. Third Int. Conf. on Quality of Service in Heterogeneous Wired/Wireless Networks, Waterloo, Ontario, Canada, August 2006, pp. 233 241 Wang, J.Z., Milstein, L.B.: CDMA overlay situations for microcellular mobile communications, IEEE Trans. Commun., 1995, 43, (2), pp. 603614 Van, G.J., Rabacy, J.: Lightweight time synchronization for sensor networks. Proc. Second ACM Int. Conf. on Wireless Sensor Networks and Applications, San Diego, Canada, September 2003, pp. 1119 Zhang, X.B., Wang, H.P., Kokhar, A.: An energy-efcient data collection protocol for mobile sensor networks. Proc. IEEE 64th Vehicular Technology Conf., Montreal, Canada, September 2006, pp. 15 Lazos, L., Radha, P., Capkun, S.: ROPE: robust position estimation in wireless sensor networks. Proc. Fourth Int. Symp. on Information Processing in Sensor Networks, Los Angeles, California, American, April 2005, pp. 324 331 Cerpa, A., Estrin, D.: ASCENT: adaptive self-conguring sensor networks topologies. Proc. 11th Joint Conf. on IEEE Computer and Communications Societies, New York, American, June 2002, pp. 12781287

14

Conclusions
15 16 17 18

In this paper, we proposed an MBC protocol for WSNs with mobile nodes. The proposed clustering protocol allows a sensor node to elect itself as a cluster-head based on its residual energy and mobility. A non-cluster-head node aims at its link stability with a cluster head during clustering according to the estimated connection time. Each noncluster-head node is allocated a timeslot for data transmission in an ascending order in a TDMA schedule based on the estimated connection time. The simulation results show that the MBC protocol outperforms both the LEACH-mobile protocol and the CBR protocol in terms of successful packet delivery rate, average energy consumption and average control overhead, and can better adapt to a highly mobile environment.

19

Acknowledgments

Appendix

The authors wish to thank Professor Jun Zheng for his valuable suggestions to improve the quality of our paper. This work was supported in part by the National Natural Science Foundation of China (No. 60872004); the Research Fund of National Mobile Communications Research Laboratory, Southeast University, under Grant No. 2010A08; the Provincial Natural Science Foundation of Hunan under Grant No. 07JJ6120 and the Excellence Youth Scientic Research Foundation of Hunan Provincial Education Department under Grant No. 08B012.

The whole protocol can be described by Algorithm 1 in Fig. 12 and Algorithm 2 in Fig. 13.

References

1 Misra, S., Reisslein, M., Xue, G.: A survey on multimedia streaming in wireless sensor networks, IEEE Commun. Surv. Tutor., 2008, 10, (4), pp. 1839 2 Lim, K.W., Ko, Y.B.: Multi-hop data harvesting in vehicular sensor networks, IET Commun., 2010, 4, (7), pp. 768 775 3 Ekici, E., Yaoyao, G., Bozdag, D.: Mobility-based communication in wireless sensor networks, IEEE Commun. Mag., 2006, 44, (7), pp. 5662 4 Felemban, E., Liu, C.G., Ekici, E.: MMSPEED: multi-path multi-speed protocol for QoS guarantee of reliability and timeliness in wireless sensor networks, IEEE Trans. Mob. Comput., 2006, 5, (6), pp. 738754 5 Zhang, Y., Dai, H.: Energy-efciency and transmission strategy selection in cooperative wireless sensor networks, J. Commun. Netw., 2007, 9, (4), pp. 1 9 6 Yuan, Y., Chen, M., Kwon, T.: A novel cluster-based cooperative MIMO scheme for multi-hop wireless sensor networks, EURASIP J. Wirel. Commun. Netw., 2006, (2), pp. 38 47 7 Heinzelman, W., Chandrakasan, A., Balakrishnan, H.: An applicationspecic protocol architecture for wireless microsensor networks, IEEE Trans. Wirel. Commun., 2002, 1, (4), pp. 660 670 8 Shu, T., Krunz, M., Vrudhula, S.: Joint optimization of transmit powertime and bit energy efciency in CDMA wireless sensor networks, IEEE Trans. Wirel. Commun., 2006, 5, (11), pp. 3109 3118 9 Liu, Y., Xiong, N., Zhao, Y., Vasilakos, A.V.: Multi-layer clustering routing algorithm for wireless vehicular sensor networks, IET Commun., 2010, 4, (7), pp. 810 816 10 Lin, K., Wang, L., Li, K.: Multi-Attribute data fusion for energy equilibrium routing in wireless sensor networks, KSII Trans. Internet Inf. Syst., 2010, 4, (1), pp. 524 11 Kim, D.S., Chung, Y.J.: Self-organization routing protocol supporting mobile nodes for wireless sensor network. Proc. First Int. MultiSymp. on Computer and Computational Sciences, Hangzhou, China, June 2006, pp. 622 626 12 Santhosh, G.K., Vinu, P.M., Poulose, K.J.: Mobility metric based LEACH-mobile protocol. Proc. 16th Int. Conf. on Advanced IET Wirel. Sens. Syst., 2011, Vol. 1, Iss. 1, pp. 3947 doi: 10.1049/iet-wss.2010.0084

Fig. 12 Algorithm 1 for non-cluster-head nodes

Fig. 13 Algorithm 2 for cluster-head nodes


47

& The Institution of Engineering and Technology 2011

Anda mungkin juga menyukai