TY - JOUR
AU - Joseph, K Suresh
AB - Abstract In a clustered cognitive radio sensor network (CRSN), the available free channels should be considered during cluster head (CH) election. Energy consumption desires have to be taken into consideration in order to improve the fairness. To meet these goals, this paper proposes a Particle Swarm Optimization-based Energy Efficient Channel Assignment (PSOEECA) technique for clustered CRSN. The election of CHs is based on the number of free channels and residual energy. At the beginning of each frame, the elected CH broadcasts a beacon signal about the selected operation mode to the cluster members (CMs). The CH performs channel assignment based on the predicted residual energy of the CMs using PSO technique. In the inter-cluster channel assignment, for transmission of data across multiple clusters, each CH exchanges the assigned channel information among them for scheduling transmission and ensuring the correctness of data delivered to the sink. Simulation results show that the proposed technique enhances throughput, network lifetime and minimizes delay and overhead. 1. INTRODUCTION Sensor nodes equipped with cognitive radio (CR) form a cognitive radio sensor network (CRSN). These wireless CRSNs are capable of sensing event signals and transmitting the readings over to the sink in a multi-hop approach through the available wireless channels. For this, the CRSN node has to sense the environment as well as the spectrum bands. The sink may or may not be CR enabled [1]. The dynamic spectrum access technology of CR allows sharing the wireless spectrum to improve the network performance. In CR, the licensed users are called primary users (PUs) and the users trying opportunistic channel access are called secondary users (SUs).The SUs have to detect the presence of a PU’s signal by means of channel sensing. Hence the SUs face the challenges of channel sensing time, channel sensing period, channel switching time and data transmission time [2]. Apart from these challenges, the SUs face additional issues like channel heterogeneity, channel quality, control channel assignment and transmission channel assignment [3, 4]. In centralized data transmission, more energy is consumed since each sensor node transmits the data to the sink which is at a larger distance. In hierarchical or cluster-based approach, data are transmitted to the sink through the selected cluster heads (CHs). Hence energy is balanced among the CHs. Clustering not only reduces the energy consumption but also extends the network lifetime and provides scalability [5]. Generally, the channel assignment (CA) problem is equivalent to the generalized graph-coloring problem, which is a well-known NP-hard problem [6]. Its NP-Hardness is proven in [7]. The challenges of CRSN motivate the design of an energy efficient CA technique for clustered CRSN [8–10]. 2. RELATED WORKS Pei et al. [11] exploited an unequal clustering method, which works adaptively for balancing energy consumption and exploiting spectrum utilization under multiple hops CRSN and achieved fairness among the CHs by designing a balanced power consumption model. Furthermore, channel availability-based CH selection was proposed to improve the stability of cluster formation. Eletreby et al. [12] proposed a spectrum and energy aware modification of LEACH [] clustering protocol. The CH selection is performed in a probabilistic manner and is performed with limited number of control message exchange. The number of channels sensed vacant is calculated as a weight in deciding the final CHs. Phuong et al. [13] proposed a power control algorithm for CRSNs, which ensures the cumulative SU interference caused to each PU is restricted to a predefined limit through local computation of each SU transmission power. Li et al. [14] proposed a framework for cluster-based energy aware CA, which aims at maximizing the network lifetime in CRSNs by balancing the energy depletion among the nodes. The predicted residual energy of the sensor nodes are used to allocate the transmission slots. Correia et al. [15] proposed a CRSN based distributed framework for development and testing of MAC protocols in a multichannel radio environment. Moreover, the best channel selection by the MAC protocols, CogTMAC and AHPTMAC, considers the application requirement, namely noise levels, transmission latency, signal interference noise rate (SINR) and received signal strength indicator (RSSI). Spachos et al. [16] proposed an opportunistic routing technique, which performs dynamic path and channel selection with quick adaptation to network changes in CRSN. The channel model used was realistic and could perform optimal signal strength evaluation with acceptable complexity. The energy resource and the computational complexity of the sensor node were considered during route selection. Akbari et al. [17] exploited the cooperative spectrum sensing technique to remove false data about the occupancy of a particular channel. The centralized controller uses a simple combination scheme to decide the presence of a PU on a particular channel. Jeong et al. [18] proposed an energy efficient channel management scheme with adaptive selection of operation mode in clustered CRSNs. The CR-based functionalities such as sensing the channel and switching over the sensed channels were given considerable attention to attain reduced depletion of energy and also ensuring minimal PU interference. 3. PROBLEM DEFINITION AND OBJECTIVES Energy efficient CA for cluster-based CRSN is proposed in [14] and [18]. But both the approaches assume existing traditional clustering techniques which did not consider the available free channels for CH selection. In [14], the R-coefficient is developed to estimate the predicted residual energy using the sensor information and channel conditions. An optimization-based CA technique based on the R-coefficient is used for intra-cluster CA. However, the issues of fairness and inter-cluster CA are mentioned as their future work. In the channel management scheme of [18], the operation mode is adaptively selected in which the channel is assigned using the traditional energy detection technique. While analyzing the existing clustering techniques on CRSN, LEAUCH [11] is the latest one which considers the number of free channels and energy, in selecting the CHs. Hence, the main aim of this research work is to design an energy efficient CA technique for clustered CRSNs with the following objectives: The CH should be selected based on energy and available channels. The CA should be done for each operation mode of CRSN. The CA should be done based on the predicted residual energy. Both intra-cluster and inter-cluster CA should be handled. The CA should be accurate and should consume less time. 4. PROPOSED METHODOLOGY 4.1. Overview In this paper, we propose a cluster-based energy efficient CA technique for CRSN. In this technique, the CH is elected based on the number of free channels and energy. The operation mode is selected by the CH at the start of each frame and the mode selected is informed through the beacon signals to the associated Cluster Members (CMs) of the cluster. As described in [14], in order to achieve total residual energy maximization model in CA, an efficient optimization technique is needed. Particle swarm optimization (PSO) has the same effectiveness as the genetic algorithm (GA) but with significantly better computational efficiency [19, 20]. Hence the CH performs the CA based on the predicted residual energy of the CMs using PSO. In inter-cluster CA, if a data is to be transmitted across multiple clusters, each CH exchanges the assigned channel information among them so that the transmission is scheduled ensuring the correctness of data delivered to the sink. Figure 1 shows the block diagram of the proposed methodology. FIGURE 1. View largeDownload slide Block diagram of proposed methodology. FIGURE 1. View largeDownload slide Block diagram of proposed methodology. 4.2. Estimation of metrics 4.2.1. Expected energy consumption The expected energy consumption for the CM i while transmitting on channel j, is estimated using the Equation (1). Eex.ij=∑z=1ZEi(z)Pjz+Ei(Z)Pjsuccess (1) where Pjz indicates the sensor i’s probability to transmit on the jth channel for z transmission slots. Pjsuccess indicates the successful transmission probability on the jth channel for a packet of length L. Ei(z) represents the sensor i’s energy consumption for transmission on z slots, which is given by Equation (2). Ei(z)=(Ecir+λdi2.z) (2) where Ecir is the energy consumed for the radio circuit, the required amplifier energy at the receiver is indicated by λ and the distance between the CM and the CH is indicated by d. 4.2.2. Residual energy model Equation (3) represents the energy consumed for transmission of z bit data by the transmitter over d distance. Etx(z,d)=Etx.e(z)+Etx.a(z,d)={z.Ee+zλtd2,d<d0z.Ee+zλad4,d≥d0 (3) where Etx.e = electronics energy, Etx.a = amplifier energy, Ee = per bit energy consumption to run the transmitter or receiver. λtd2 and λad4 are the amplifier energy that depends on the transmitter distance. The receiver energy consumption is represented as per Equation (4). Erx(z)=Erx.e(z)=zEe .(4) The estimated residual energy after z bit data transmission by each node is given as per Equation (5). Eres=[Ei−(Etx(z,d)+Erx(z))] (5) where the initial residual energy of each node is indicated by Ei. 4.3. Clustering using LEAUCH protocol The steps involved in LEAUCH [11] clustering are presented in Algorithm 1. Notations used in the clustering algorithm with their descriptions are summarized in Table 1. Algorithm 1 LEAUCH clustering Each ki broadcasts REQ to all its neighbors with same Ptx; if node kj receives REQthen  It forms QKi such that  QKi = {kj| kj is candidate CH, d(ki, kj) < max(Wki, Wkj)}; end ifEres(kj) = Max (Eres(ki)), kj ∈QKi, then  kj becomes CH and broadcasts its decision to its neighbor nodes for competition;  else ifki receives the winning message from kj (kj∈QKi) then   ki quits the competition and broadcast the quit message;  end  else ifki receives the quit message from kj (kj ∈ QKi) then   ki removes kj from the set of its adjacent CHs;  end end ki broadcasts <Nid, CAlist> after winning the election; ifkj receives the message then  ifNCidle >0 then   kj sends a REP to ki to join the cluster;  end end ki records <Nid, CAlist> of kj; ki selects the common channel which is shared by most of the nodes and broadcasts it; Each ki broadcasts REQ to all its neighbors with same Ptx; if node kj receives REQthen  It forms QKi such that  QKi = {kj| kj is candidate CH, d(ki, kj) < max(Wki, Wkj)}; end ifEres(kj) = Max (Eres(ki)), kj ∈QKi, then  kj becomes CH and broadcasts its decision to its neighbor nodes for competition;  else ifki receives the winning message from kj (kj∈QKi) then   ki quits the competition and broadcast the quit message;  end  else ifki receives the quit message from kj (kj ∈ QKi) then   ki removes kj from the set of its adjacent CHs;  end end ki broadcasts <Nid, CAlist> after winning the election; ifkj receives the message then  ifNCidle >0 then   kj sends a REP to ki to join the cluster;  end end ki records <Nid, CAlist> of kj; ki selects the common channel which is shared by most of the nodes and broadcasts it; Algorithm 1 LEAUCH clustering Each ki broadcasts REQ to all its neighbors with same Ptx; if node kj receives REQthen  It forms QKi such that  QKi = {kj| kj is candidate CH, d(ki, kj) < max(Wki, Wkj)}; end ifEres(kj) = Max (Eres(ki)), kj ∈QKi, then  kj becomes CH and broadcasts its decision to its neighbor nodes for competition;  else ifki receives the winning message from kj (kj∈QKi) then   ki quits the competition and broadcast the quit message;  end  else ifki receives the quit message from kj (kj ∈ QKi) then   ki removes kj from the set of its adjacent CHs;  end end ki broadcasts <Nid, CAlist> after winning the election; ifkj receives the message then  ifNCidle >0 then   kj sends a REP to ki to join the cluster;  end end ki records <Nid, CAlist> of kj; ki selects the common channel which is shared by most of the nodes and broadcasts it; Each ki broadcasts REQ to all its neighbors with same Ptx; if node kj receives REQthen  It forms QKi such that  QKi = {kj| kj is candidate CH, d(ki, kj) < max(Wki, Wkj)}; end ifEres(kj) = Max (Eres(ki)), kj ∈QKi, then  kj becomes CH and broadcasts its decision to its neighbor nodes for competition;  else ifki receives the winning message from kj (kj∈QKi) then   ki quits the competition and broadcast the quit message;  end  else ifki receives the quit message from kj (kj ∈ QKi) then   ki removes kj from the set of its adjacent CHs;  end end ki broadcasts <Nid, CAlist> after winning the election; ifkj receives the message then  ifNCidle >0 then   kj sends a REP to ki to join the cluster;  end end ki records <Nid, CAlist> of kj; ki selects the common channel which is shared by most of the nodes and broadcasts it; TABLE 1. Notations used in clustering algorithm. Notations Description ki,kj Nodes QKi Set of adjacent CHs of ki W Competition radius REQ Request message REP Reply message Ptx Transmit power level CAlist Channel availability list NCidle Number of idle channels available Nid Node id Notations Description ki,kj Nodes QKi Set of adjacent CHs of ki W Competition radius REQ Request message REP Reply message Ptx Transmit power level CAlist Channel availability list NCidle Number of idle channels available Nid Node id TABLE 1. Notations used in clustering algorithm. Notations Description ki,kj Nodes QKi Set of adjacent CHs of ki W Competition radius REQ Request message REP Reply message Ptx Transmit power level CAlist Channel availability list NCidle Number of idle channels available Nid Node id Notations Description ki,kj Nodes QKi Set of adjacent CHs of ki W Competition radius REQ Request message REP Reply message Ptx Transmit power level CAlist Channel availability list NCidle Number of idle channels available Nid Node id In Algorithm 1, each node ki broadcasts a request message (REQ) with the same power to its neighbor nodes. The format of a request message is given in Table 2. TABLE 2. Format of REQ message. Nid W Eres Nid W Eres Note: In order to save energy, the competition radius (W) is set to R0. It reveals the range up to which the message can be broadcasted. TABLE 2. Format of REQ message. Nid W Eres Nid W Eres Note: In order to save energy, the competition radius (W) is set to R0. It reveals the range up to which the message can be broadcasted. After receiving REQ message, each node makes a decision to announce itself as a CH or not after analyzing its adjacent CH set. The node ki with the highest Eres in QKi becomes the CH and broadcasts its decision to its neighboring nodes for competition. After decision on the final CHs, each non-participated node in the election switches from the sleep mode to active mode and merges itself with a neighboring cluster. The node ki broadcasts its message after winning the election that includes Nid and CAlist. When node kj, which is in the transmission range of ki receives the message, it verifies whether it contains one or more element with the idle channels. If it contains, then it will send a reply message to ki to join the cluster. Furthermore, ki records the requesting nodes’ IDs and the list of available channels which are further used for managing cluster-based communication. The remaining non-clustered nodes join a nearby cluster in which the associated CH of the cluster demands minimum energy consumption for communication. Figure 2 demonstrates the cluster formation phase. CH1, CH2, CH3 and CH4 are selected as CHs as per the above algorithm. FIGURE 2. View largeDownload slide Cluster formation. FIGURE 2. View largeDownload slide Cluster formation. 4.4. Operation mode selection The operation modes [18] of each node are categorized into data transmission/reception (DTR), sense operating channel/backup channel (SO/SB) and change operating channel/backup channel (CO/CB). The CH performs adaptive selection of one of the five operating modes based on each mode’s sensing outcome and energy consumption, and further broadcasts the decision on the selected mode at the start of each time frame through the beacon. The processes involved in each mode are as follows: SO (or SB) Mode During SO mode, the CMs are directed by the CH (including itself) for operating channel sensing. The instructed CMs send the feedback on sensing to the CH. The CH estimates the next operating mode based on the operating channel state. During SB Mode, a new backup channel is randomly selected by the CH and the CMs are instructed to perform sensing on the selected channel. Similar to SO mode, CH estimates the new backup channel state with the sensed feedback. Based on the estimated new backup channel state, the selection of the subsequent mode of operation is decided. DTR Mode A beacon containing the transmission schedule is broadcasted by the CH to the CMs during the DTR mode. Each CM follows the transmission schedule in sending the data to the CH and switches to sleep state to save energy. The CH also switches to sleep state after finishing data reception from the CMs. The CH switches to wake up state at the end of DTR mode and the next mode of operation is decided by the CH. CO (or CB) mode The CH and the associated CMs senses the newly selected operating channel (previously selected as backup channel) and randomly selected new backup channel. Once the network commences its operation on the new operating channel, the CH directs all the CMs to synchronize to the new channel through a beacon signal. After reporting the sensed results of both the channels, each CM joins the channel and intimates the CH with a joining message. The CH acknowledges the join message with a response message. The next mode of operation is decided by the CH on the basis of the gathered sensed information. Note: The network operates on the new operating channel if the next mode of operation is not CO. 4.5. Particle swarm optimization PSO is inspired by social behavior of bird flocking and used as a population based stochastic optimization technique. In PSO, the particle or potential solutions fly through the problem space by following the fitter members of the swarm (current optimum particles). PSO demands simple implementation with lesser number of individuals. Also, with smaller time steps, PSO demands lower computational cost and fewer parameter adjustments than other evolutionary techniques. The PSO technique self-updates the particles by tracking the following two variables: Gbest(i): Global best indicates the particle with the highest proximity to the target value. Pbest(i): Personal best indicates the particle’s highest proximity ever achieved to the target value. Equation (6) depicts the new velocity and new location of the particle which are updated after updating the above two values. λi(t+1)=Ω×λid(t)+L1×rand()×[Pbest(t)−σid(t)]+L2×rand()×[Gbest(t)−σid(t)] (6) where σid(t+1)=σid(t)+λid(t+1), Number of initial particles: D; 1≤i≤D, Search space dimension: V; 1≤d≤V, Maximum number of iterations: Dmax; 1≤t≤Dmax, Inertia weight: Ω, Learning factor: L1, L2, Random number in the range {0, 1}: rand () According to an objective fitness function, the optimal position and optimal location of an individual are updated. Equation (7) represents the global and individual optimal values. Pbest(i)(t+1)={Pbest(i)(t),iffitness(σid(t+1))≥fitness(Pbest(i)(t))σid(t+1),iffitness(σid(t+1))<fitness(Pbest(i)(t)). (7) 4.6. Channel assignment Let i be the channel and for the channel detected idle in the current time slot, the probability to move to a busy state in the next time slot is Xi. For the channel detected busy in the current time slot, the probability to move to an idle state in the next time slot is Yi. Based on the R-coefficient, the CH performs the CA. During the SO mode, the PU activity level on a channel is determined by the CM based on the following probabilities: Pidle: It is the probability that the channel is idle and is derived using the steady state probabilities as given in Equation (8). Pidle=YiXi+Yi (8) Psuccess: A packet’s successful transmission probability. It is derived using Equation (9). Psuccess=P(idlefornextZslots/initiallyidle)=(1−Xi)Z (9) For Z transmission slots, the probability of the channel remaining idle is represented using Equation (9). CM forwards the Psuccess and Pidle to CH. CH then estimates the R-coefficient for each pair of sensor i and channel j. The optimum pair (i,j) is then determined using PSO technique in which the fitness function is formed in terms of the R-coefficient. 4.6.1. PSO-based optimum node-channel pair estimation The steps involved in PSO-based optimum node- channel pair are as follows: The network is initialized with the swarm particles (SP) with random position for each particle. Each node’s position and velocity (λi, σi) is equivalent to a search window represented by a SPi. The node’s parameters, namely expected energy consumption (Eex) and current residual energy (Eres) (Estimated in Sections 4.2.1 and 4.2.2) is monitored by each SPi. The fitness function (Fi) of each particle based on the parameters monitored is given as per Equation (10). Fi=(μ1×Eres)−(μ2×Eex) (10) where μ1 and μ2 are the controlling factors with values, μ1 = 0.6 and μ2 = 0.4. The global best (Gbest(i)) and personal best (Pbest(i)), and each particle’s position are calculated (described in Section 4.5). Each particle includes the information given in Table 3. Each particle’s position and velocity are updated as per Equation (6). Update the position of Pbest(i) and Gbest(i) based on the following condition: if  Fi>Fi(Pbest(i))then  According to the fitness value Fi, the position of Pbest(i) is updated; end if  Fi>Fi(Gbest(i)) then  According to the fitness value Fi, the position of Gbest(i) is updated; end The best pair is considered to be the one with the global best value. If still, some channels are remaining, then CH changes the mode to SB in which backup channels are assigned to the nodes with low R-coefficient values. TABLE 3. Particle information. Position Velocity Personal best Fitness λ1,λ2,..,λM σ1,σ2,..,σM Pbest(i) Fi Position Velocity Personal best Fitness λ1,λ2,..,λM σ1,σ2,..,σM Pbest(i) Fi TABLE 3. Particle information. Position Velocity Personal best Fitness λ1,λ2,..,λM σ1,σ2,..,σM Pbest(i) Fi Position Velocity Personal best Fitness λ1,λ2,..,λM σ1,σ2,..,σM Pbest(i) Fi 4.6.2. Complexity of the algorithm The complexity of PSOEECA algorithm in all iterations could be computed according to Equation (11). O((m×g)+(h×g)) (11) where m: problem dimension, g: size of population and h: objective function cost. Moreover, s/g iterations are performed by the PSOEECA algorithm, where the upper limit on the number of evaluations allowed on the objective function is represented as s. Therefore, the complexity of PSOEECA, which is controlled by the number of executed evaluations and the objective function evaluation cost, could be derived according to Equation (12). O((m×s)+(h×s)) .(12) 4.6.3. Mode change during throughput degradation During data transmission, there may be either PU activity or throughput degradation for any CM. During this scenario, the CH performs the following process: Changes the mode to CO. CM on receiving mode change information from CH senses the activity level of the other channels and reports to the CH. The CH then chooses another channel based on the R-coefficient. 4.7. Inter-cluster CA The sequence of steps involved in inter-cluster CA is presented in Algorithm 2. Algorithm 2 Inter-cluster CA Let CMij be the CM of CH CHi; Let R1, R2, …, Rk be the routes of k members CMij to the sink; for each CHi, i = 1, 2…. do  Store the routes R1, R2, …, Rk of its members CMij;  if S ∈ CHithen   CHi fetches the route Rs of source S;   if D ∉ CHithen    CHi look up the target CHD from Rs;    CHi transmits the channel information of S to CHD;    CHD transmits the assigned channel information of its members CMDj;    S transmits the data to CHi using the channel assigned by CHi using PSOEECA;    CHi transmits the data to CHD using its free channel without any conflict to the channels of CHD;    CHD transmits the data to the sink using its free channel without any conflict to the channels of CHi;   end  end end Let CMij be the CM of CH CHi; Let R1, R2, …, Rk be the routes of k members CMij to the sink; for each CHi, i = 1, 2…. do  Store the routes R1, R2, …, Rk of its members CMij;  if S ∈ CHithen   CHi fetches the route Rs of source S;   if D ∉ CHithen    CHi look up the target CHD from Rs;    CHi transmits the channel information of S to CHD;    CHD transmits the assigned channel information of its members CMDj;    S transmits the data to CHi using the channel assigned by CHi using PSOEECA;    CHi transmits the data to CHD using its free channel without any conflict to the channels of CHD;    CHD transmits the data to the sink using its free channel without any conflict to the channels of CHi;   end  end end Algorithm 2 Inter-cluster CA Let CMij be the CM of CH CHi; Let R1, R2, …, Rk be the routes of k members CMij to the sink; for each CHi, i = 1, 2…. do  Store the routes R1, R2, …, Rk of its members CMij;  if S ∈ CHithen   CHi fetches the route Rs of source S;   if D ∉ CHithen    CHi look up the target CHD from Rs;    CHi transmits the channel information of S to CHD;    CHD transmits the assigned channel information of its members CMDj;    S transmits the data to CHi using the channel assigned by CHi using PSOEECA;    CHi transmits the data to CHD using its free channel without any conflict to the channels of CHD;    CHD transmits the data to the sink using its free channel without any conflict to the channels of CHi;   end  end end Let CMij be the CM of CH CHi; Let R1, R2, …, Rk be the routes of k members CMij to the sink; for each CHi, i = 1, 2…. do  Store the routes R1, R2, …, Rk of its members CMij;  if S ∈ CHithen   CHi fetches the route Rs of source S;   if D ∉ CHithen    CHi look up the target CHD from Rs;    CHi transmits the channel information of S to CHD;    CHD transmits the assigned channel information of its members CMDj;    S transmits the data to CHi using the channel assigned by CHi using PSOEECA;    CHi transmits the data to CHD using its free channel without any conflict to the channels of CHD;    CHD transmits the data to the sink using its free channel without any conflict to the channels of CHi;   end  end end During inter-cluster CA, CH performs the following process: It stores the entire route to the base station for each of its members. If a data is to be transmitted across multiple clusters, each CH exchanges the assigned channel information among them, so that the transmission is scheduled ensuring the correctness of data delivery to the base station. The process flow diagram for PSOEECA is depicted in Fig. 3. FIGURE 3. View largeDownload slide Process Flow Diagram. FIGURE 3. View largeDownload slide Process Flow Diagram. 5. SIMULATION RESULTS 5.1. Simulation parameters We use NS-2 to simulate our proposed PSOEECA technique. The simulation settings and parameters are summarized in Table 4. The number of nodes and area size are fixed with respect to the transmission range. The propagation model and antenna are considered based on the basic IEEE 802.11 standards. The MAC protocol and the number of available channels are fixed as per the cognitive radio network simulator for NS-2. TABLE 4. Simulation parameters. Number of nodes 50, 100, 150 and 200 Primary users 5 Area size 1000 × 1000 m Transmission range 75 m MAC protocol IEEE 802.22 contention based MAC Simulation time 50 s Traffic source Constant bit rate Transmission rate 200 Kb/s to 1000 Kb/s Packet size 500 bits Available number of channels 2 to 10 Propagation TwoRayGround Antenna OmniAntenna Initial energy 10 Joules Transmission power 0.8 watts Receiving power 0.3 watts Number of nodes 50, 100, 150 and 200 Primary users 5 Area size 1000 × 1000 m Transmission range 75 m MAC protocol IEEE 802.22 contention based MAC Simulation time 50 s Traffic source Constant bit rate Transmission rate 200 Kb/s to 1000 Kb/s Packet size 500 bits Available number of channels 2 to 10 Propagation TwoRayGround Antenna OmniAntenna Initial energy 10 Joules Transmission power 0.8 watts Receiving power 0.3 watts TABLE 4. Simulation parameters. Number of nodes 50, 100, 150 and 200 Primary users 5 Area size 1000 × 1000 m Transmission range 75 m MAC protocol IEEE 802.22 contention based MAC Simulation time 50 s Traffic source Constant bit rate Transmission rate 200 Kb/s to 1000 Kb/s Packet size 500 bits Available number of channels 2 to 10 Propagation TwoRayGround Antenna OmniAntenna Initial energy 10 Joules Transmission power 0.8 watts Receiving power 0.3 watts Number of nodes 50, 100, 150 and 200 Primary users 5 Area size 1000 × 1000 m Transmission range 75 m MAC protocol IEEE 802.22 contention based MAC Simulation time 50 s Traffic source Constant bit rate Transmission rate 200 Kb/s to 1000 Kb/s Packet size 500 bits Available number of channels 2 to 10 Propagation TwoRayGround Antenna OmniAntenna Initial energy 10 Joules Transmission power 0.8 watts Receiving power 0.3 watts The simulation topology is shown in Fig. 4. In Fig. 4, the CHs are marked in yellow, gateway or relay nodes are marked in red and the sink is marked in blue color. FIGURE 4. View largeDownload slide Simulation topology. FIGURE 4. View largeDownload slide Simulation topology. 5.2. Performance metrics The proposed PSOEECA technique is compared with the LEAUCH [11] protocol and residual energy aware CA (REACA) technique [14]. Data delivery delay: It is the time a packet takes to get transmitted across a network from source to destination. Average packet delivery ratio: It is the ratio of the number of packets received successfully and the total number of packets transmitted. Packet drop: It is the number of packets dropped during the data transmission. Throughput: The throughput is the amount of data that can be sent from the sources to the destination. Residual energy: It is the amount of energy remains in the nodes after the data transmission. Control overhead: The ratio of control packets sent to the total number of packets sent. Network lifetime: The amount of time taken for a sensor network to be fully operative. 5.3. Results and analysis The results of simulation are presented in this section. 5.3.1. Varying the number of nodes In order to test the scalability of the schemes, the number of nodes is varied as 50, 100, 150 and 200. Figure 5 shows the data delivery delay involved in the three approaches. The delay increases as the number of nodes are increased since the clustering time increases. As it can be seen from Fig. 5, PSOEECA has the least delay in the range of 0.11–0.4 s, when compared to LEAUCH and REACA approaches. As the PSO technique selects the most optimal node and channel pair in considerably less iteration, the delay of PSOEECA is 37% lesser than REACA. As far as LEAUCH is considered, the delay of PSOEECA is 82% less since LEAUCH does not perform optimal CA. FIGURE 5. View largeDownload slide Data delivery delay versus number of nodes. FIGURE 5. View largeDownload slide Data delivery delay versus number of nodes. Figures 6 and 7 show the results of the average packet delivery ratio and packet drop for the three approaches. The increase in the number of nodes results in more channels being engaged so that the number of packet drops increases. As seen from Figs 6 and 7, PSOEECA has the least packet drops around 217 and has the highest delivery ratio around 0.99, when compared to LEAUCH and REACA approaches. The use of PSO-based CA increases the correctness of data delivery in PSOEECA. Hence it has 87% lesser drop and 12% higher delivery ratio, than REACA. Similarly, it has 92% lesser drop and 18% higher delivery ratio than LEAUCH. FIGURE 6. View largeDownload slide Packet delivery ratio versus number of nodes. FIGURE 6. View largeDownload slide Packet delivery ratio versus number of nodes. FIGURE 7. View largeDownload slide Packet drop versus number of nodes. FIGURE 7. View largeDownload slide Packet drop versus number of nodes. Figure 8 shows the results of throughput obtained for the three approaches. When the number of nodes is increased, throughput slightly decreases for LEAUCH and REACA, whereas it remains almost constant for PSOEECA. As seen from the Fig. 8, PSOEECA has the highest throughput around 0.5Mbps, when compared to LEAUCH and REACA approaches. The use of PSO-based CA increases the correctness of data delivery in PSOEECA. Hence it has 39% higher throughput than REACA and 30% higher throughput than LEAUCH. FIGURE 8. View largeDownload slide Throughput versus number of nodes. FIGURE 8. View largeDownload slide Throughput versus number of nodes. The average residual energies for the three approaches are depicted in Fig. 9. As the number of nodes is increased, the average residual energy of nodes decreases for all the three techniques. Both LEAUCH and REACA, have long channel sensing periods. Since PSOEECA efficiently switches the operating and backup channels, it has less channel sensing periods. Hence the average residual energy of PSOEECA is 25% more than REACA and 22% more than LEAUCH. FIGURE 9. View largeDownload slide Residual energy versus number of nodes. FIGURE 9. View largeDownload slide Residual energy versus number of nodes. Figure 10 shows the control overhead obtained for all the three approaches. The control overhead is measured as the ratio of control packets sent to the total number of packets sent. Since REACA does not possess an efficient clustering scheme, the overhead involved in the cluster formation will be more. Hence it has the highest overhead when compared to LEAUCH and PSOEECA. However, the channel establishing cost will be slightly more in the case of LEAUCH. So as expected, PSOEECA has the least overhead than the other two approaches. It has 35% lesser overhead than REACA and 24% lesser overhead than LEAUCH. FIGURE 10. View largeDownload slide Control overhead versus number of nodes. FIGURE 10. View largeDownload slide Control overhead versus number of nodes. Figure 11 shows the average network lifetime for all the three approaches. Since the average residual energy of nodes decreases as shown in Fig. 9, the network lifetime will also tend to decrease, as the number of nodes is increased. However, PSOEECA has the highest lifetime than the other two approaches. It has 27% higher lifetime than REACA and 21% higher lifetime than LEAUCH. FIGURE 11. View largeDownload slide Network lifetime versus number of nodes. FIGURE 11. View largeDownload slide Network lifetime versus number of nodes. 5.3.2. Varying the transmission rate In order to test the effect of various loads, the transmission rate of the CBR traffic is varied from 200Kb/s to 1000 Kb/s for 200 nodes. Figure 12 shows the data delivery delay involved in the three approaches. The increase in transmission rate increases the delay and causes the transmission time to increase. As it can be seen from the Fig. 12, PSOEECA has the least delay in the range of 0.3 to 1.2 s, when compared to LEAUCH and REACA approaches. As the PSO technique selects the most optimal node and channel pair in considerably less iteration, the delay of PSOEECA is 74% lesser than REACA. As far as LEAUCH is considered, the delay of PSOEECA is 90% less since LEAUCH does not perform optimal CA. FIGURE 12. View largeDownload slide Data delivery delay versus transmission rate. FIGURE 12. View largeDownload slide Data delivery delay versus transmission rate. Figures 13 and 14 show the results of average packet delivery ratio and packet drop for the three approaches. The increase in the number of nodes results in more channels being engaged so that the number of packet drops increases. As seen from the Figs 13 and 14, PSOEECA has the least packet drops around 3020 and has the highest delivery ratio around 0.84, when compared to LEAUCH and REACA approaches. The use of PSO-based CA increases the correctness of data delivery in PSOEECA. Hence it has 36% lesser drop and 59% higher delivery ratio, than REACA. Similarly, it has 72% lesser drop and 46% higher delivery ratio than LEAUCH. FIGURE 13. View largeDownload slide Delivery ratio versus transmission rate. FIGURE 13. View largeDownload slide Delivery ratio versus transmission rate. FIGURE 14. View largeDownload slide Packet drop versus transmission rate. FIGURE 14. View largeDownload slide Packet drop versus transmission rate. Figure 15 shows the results of throughput obtained for the three approaches. As seen from the Fig. 15, PSOEECA has the highest throughput around 3.0Mbps, when compared to LEAUCH and REACA approaches. The use of PSO-based CA increases the correctness of data delivery in PSOEECA. Hence it has 64% higher throughput than REACA and 66% higher throughput than LEAUCH. FIGURE 15. View largeDownload slide Throughput versus transmission rate. FIGURE 15. View largeDownload slide Throughput versus transmission rate. The average residual energies for the three approaches are depicted in Fig. 16. Both LEAUCH and REACA have long channel sensing periods. Since PSOEECA efficiently switches the operating and backup channels, it has less channel sensing periods. Hence the average residual energy of PSOEECA is 17% more than REACA and 24% more than LEAUCH. FIGURE 16. View largeDownload slide Residual energy versus transmission rate. FIGURE 16. View largeDownload slide Residual energy versus transmission rate. Figure 17 shows the control overhead obtained for all the three approaches when the transmission rate is varied. Since REACA does not possess an efficient clustering scheme, the overhead involved packet forwarding will be more. Hence it has the highest overhead when compared to LEAUCH and PSOEECA. However, the channel establishing cost will be slightly more in the case of LEAUCH. So as expected, PSOEECA has the least overhead than the other two approaches. It has 47% lesser overhead than REACA and 41% lesser overhead than LEAUCH. FIGURE 17. View largeDownload slide Control overhead versus transmission rate. FIGURE 17. View largeDownload slide Control overhead versus transmission rate. 6. CONCLUSION In this paper, we have proposed a cluster-based energy efficient CA technique for CRSN. In this technique, the CH is elected based on the number of free channels and energy. At the beginning of each frame, the elected CH selects the operation mode and informs the CMs of the selected mode through beacon signals. The CH performs the CA based on the predicted residual energy of the CM using PSO. During inter-cluster CA, if data are to be transmitted across multiple clusters, all CHs exchange the assigned channel information among them so that the transmission is scheduled ensuring the correctness of data delivery to the sink. By simulation results, we showed that the proposed technique enhances throughput and minimizes energy consumption, delay and overhead. REFERENCES 1 Prasain , P. , Seong-Ro , L. , Sun-Kok , N. and Dong-You , C. ( 2014 ) Issues in realization of cognitive radio sensor networks . Int. J. Control Autom. , 7 , 141 – 152 . Google Scholar CrossRef Search ADS 2 Joshua , A. , Nurul , M.A.L. , Sharifah , K.S.Y. and Norsheila , F. ( 2015 ) Channel decision in cognitive radio enabled sensor networks: a reinforcement learning approach . Int. J. Eng. Technol. , 7 , 1394 – 1404 . 3 Hyung , S.K. , Waleed , E. , Khalid , A.B. , Al-Sakib , K.P. and Najam , H. ( 2014 ) Advances in cognitive radio sensor networks . Int. J. Distrib. Sensor Netw. , 10 , 1 – 3 . 4 Siew , Z.W. , Wong , C.H. , Chin , C.S. , Kiring , A. and Teo , K.T.K. ( 2012 ) Cluster Heads Distribution of Wireless Sensor Networks via Adaptive Particle Swarm Optimization. Proceedings of CICSyN 2012, Phuket, Thailand, July 24–26, pp. 78–83. IEEE. 5 Usman , M. and Muhammad , K.S. ( 2014 ) Cluster based energy efficient sensing for cognitive radio sensor networks . Int. J. Comput. Appl. , 88 , 14 – 19 . 6 Jie , W. , Ying , D. and Yanchao , Z. ( 2011 ) Local Channel Assignments in Cognitive Radio Networks, Proceedings of ICCCN 2011, Maui, HI, USA, July 31–Aug 4, pp. 1–6. IEEE. 7 Raniwala , A. , Gopalan , K. and Chiueh , T. ( 2004 ) Centralized channel assignment and routing algorithms for multi channel WMNs . ACM SIGMOBILE Mobile Comp. Commun. Rev. , 8 , 50 – 65 . Google Scholar CrossRef Search ADS 8 Latiff , N.M.A. , Tsimenidis , C.C. and Sharif , B.S. ( 2007 ) Energy-Aware Clustering for Wireless Sensor Networks using Particle Swarm Optimization. Proceedings of PIMRC 2007, Athens, Greece, September 3–7, pp. 1–5. IEEE. 9 Lazos , L. , Liu , S. and Krunz , M. ( 2009 ) Spectrum Opportunity-Based Control Channel Assignment in Cognitive Radio Networks, Proceedings of SECON 2009, Rome, Italy, June 22–26, pp. 1–9. IEEE. 10 Huang , J. , Zeng , X. , Jian , X. , Tan , X. and Zhang , Q. ( 2017 ) Opportunistic capacity-based resource allocation for chunk-based multi-carrier cognitive radio sensor networks . Sensors , 7 , 175 . Google Scholar CrossRef Search ADS 11 Errong , P. , Haozhe , H. , Zehua , S. , Bin , S. and Tianqi , Z. ( 2015 ) LEAUCH: low-energy adaptive uneven clustering hierarchy for cognitive radio sensor network . EURASIP J. Wirel. Commun. Netw. , 2015, 1 – 8 . 12 Rashad , M.E. , Hany , M.E. and Mohamed , M.K. ( 2014 ) CogLEACH: A Spectrum Aware Clustering Protocol for Cognitive Radio Sensor Networks. Proceedings of CROWNCOM 2014, Oulu, Finland, July 2–4, pp. 179–184. IEEE. 13 Tran , M.P. and Dong-Seong , K. ( 2013 ) Robust distributed power control for cognitive radio based industrial wireless sensor networks . Adv. Sci. Technol. Lett. , 41 , 9 – 12 . 14 Xiaoyuan , L. , Dexiang , W. , Janise , M. and Jianmin , C. ( 2011 ) Residual Energy Aware Channel Assignment in Cognitive Radio Sensor Networks. Proceedings of WCNC 2011, Cancun, Quintana Roo, Mexico, March 28–31, pp. 398–403. IEEE. 15 Luiz , H.A.C. , Erasmo , E.O. , Daniel , F.M. , Pedro , M.M. , Antonio , A.F.L. and Jorge , S.S. ( 2012 ) A Framework for Cognitive Radio Wireless Sensor Networks. Proceedings of ISCC 2012, Cappadocia, Turkey, July 1–4, pp. 000611–000616. IEEE. 16 Petros , S. and Dimitrios , H. ( 2014 ) Scalable dynamic routing protocol for cognitive radio sensor networks . IEEE. Sens. J. , 14 , 2257 – 2266 . Google Scholar CrossRef Search ADS 17 Mohammad , A. and Abolfazl , F. ( 2011 ) A fault-tolerant cooperative spectrum sensing algorithm over cognitive radio network based on wireless sensor network . Wirel. Sensor Netw. , 3 , 83 – 91 . Google Scholar CrossRef Search ADS 18 Jeong , A.H. , Wha , S.J. and Dong , G.J. ( 2011 ) Energy-efficient channel management scheme for cognitive radio sensor networks . IEEE Trans. Veh. Technol. , 60 , 1905 – 1910 . Google Scholar CrossRef Search ADS 19 Rania , H. , Babak , C. and Olivier , D.W. ( 2005 ) A comparison of particle swarm optimization and the genetic algorithm. Proceeedings of the First AIAA Multidisciplinary Design Optimization Specialist Conference, pp. 18 – 21 . 20 Sidhartha , P. and Narayana , P. ( 2008 ) Comparison of particle swarm optimization and genetic algorithm for FACTS-based controller design . Appl. Soft Comput. , 8 , 1418 – 1427 . Google Scholar CrossRef Search ADS Author notes Handling editor: Ing-Ray Chen © The British Computer Society 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
TI - Particle Swarm Optimization-Based Energy Efficient Channel Assignment Technique for Clustered Cognitive Radio Sensor Networks
JF - The Computer Journal
DO - 10.1093/comjnl/bxx119
DA - 2018-01-10
UR - https://www.deepdyve.com/lp/oxford-university-press/particle-swarm-optimization-based-energy-efficient-channel-assignment-nwnrHHEK7d
SP - 1
EP - 936
VL - Advance Article
IS - 6
DP - DeepDyve
ER -