5864
3, JUNE 2019
Letters
NSAC: A Novel Clustering Protocol in Cognitive Radio
Sensor Networks for Internet of Things
Meng Zheng , Member, IEEE, Si Chen, Wei Liang , Senior Member, IEEE, and Min Song
Abstract—Clustering is an effective method to manage communica-
tions in cognitive radio sensor networks (CRSNs). This letter proposes a
network stability-aware clustering (NSAC) protocol for CRSNs. Spectrum
dynamics and energy consumption are for the first time simultaneously
integrated into the protocol design of NSAC. Extensive simulations show
that the proposed NSAC protocol obviously outperforms existing methods
in the aspects of network stability and energy consumption.
Index Terms—Clustering protocol, cognitive radio, Internet of
Things, wireless sensor networks.
I. I NTRODUCTION
ITH respect to the significant growth in request for wireless
W services and the scarcity of the spectrum resources, cogni-
tive radio technology which allows cognitive sensors (CSs) to access
vacant licensed channels in an opportunistic way, has been proposed.
Cognitive radio sensor networks (CRSNs) have been widely recog-
nized as a promising solution to reliable and time-efficient wireless
sensor networks for Internet of Things [1]–[6].
Clustering is an effective way to manage communications in
CRSNs [6]. A clustered CRSN organizes the network by impos-
ing local coordinators so that spectrum-aware communications can
be efficiently formed in a structured way. However, due to the fre-
quent topology variation of CRSNs by spectrum dynamics, existing
methods [6]–[8] suffer high clustering frequency, which incurs large
communication overhead.
This letter for the first time proposes a network stability-aware clus-
tering (NSAC) protocol for CRSNs. In the NSAC protocol, each CS
broadcasts its remaining energy and available channels, and searches
Manuscript received July 22, 2018; revised October 31, 2018; accepted
February 5, 2019. Date of publication February 8, 2019; date of current version
June 19, 2019. This work was supported in part by the National Key Research
and Development Program of China under Grant 2017YFE0101300, in part by
the NSFC under Grant 61673371, Grant 71661147005, and Grant 71602124,
in part by the Liaoning Provincial NSFC under Grant 20170540662, and in
part by the Youth Innovation Promotion Association, Chinese Academy of
Sciences under Grant 2015157. (Corresponding author: Wei Liang.)
M. Zheng and W. Liang are with the State Key Laboratory of Robotics,
Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang
110016, China, and also with the Key Laboratory of Networked Control
Systems, Shenyang Institute of Automation, Chinese Academy of Sciences,
Shenyang 110016, China (e-mail: zhengmeng_6@sia.cn; weiliang@sia.cn).
S. Chen is with the State Key Laboratory of Robotics, Shenyang Institute of
Automation, Chinese Academy of Sciences, Shenyang 110016, China, with
the Key Laboratory of Networked Control Systems, Shenyang Institute of
Automation, Chinese Academy of Sciences, Shenyang 110016, China, and
also with the University of Chinese Academy of Sciences, Beijing 100049,
China (e-mail: chensi@sia.cn).
M. Song is with the College of Information, Shenyang Institute of
Engineering, Shenyang 110136, China (e-mail: songmin@sie.edu.cn).
Digital Object Identifier 10.1109/JIOT.2019.2898166
2327-4662 c 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO.
the maximum edge biclique (MEB) based on its neighbors and chan-
nel qualities. The weight of one CS is calculated based on remaining
energies, the number of CSs and the channel qualities of available
common control channels in its MEB. The CS with the largest weight
in the neighborhood will be selected as the head of one cluster, and
the rest of CSs in its MEB join the cluster as members. The clustering
process repeats until all CSs are clustered.
II. S YSTEM M ODEL
This letter considers one CRSN that consists of one sink and N
static CSs. M licensed channels are opportunistically available for
the CRSN. The CRSN can access the licensed channels only when
primary users (PUs) are inactive. The PU activity on the ith (1 ≤
i ≤ M) licensed channel can be modeled as an independent and
identically distributed random process with busy and idle states. The
probabilities of the facts that the ith licensed channel are idle and busy
are given by pi and qi . Obviously, pi + qi = 1. Let Ni and Ci denote
the sets of neighbors and available channels of CS i, respectively.
(Ni , Ci ) at different CSs may be different.
Considering the probability of idle state pi and the average idle
duration Mi , we introduce a channel quality metric Qi for the ith
licensed channel (public information) as follows:
 
Qi = 1 + logε pi Mi (1)
where ε > 1 denotes a parameter which indicates the preference on
pi (select big ε) or Mi (select small ε).
III. NSAC P ROTOCOL
A. Protocol Description
As frequent clustering incurs large communication overhead,
NSAC aims to build a stable cluster-based architecture by consid-
ering spectrum dynamics and energy consumption. The details of
NSAC are summarized in Algorithm 1.
Initialization: CS i detects its residual energy Ei and the set of
available channels Ci . All (Ei , Ci ) are broadcast.
Weight Calculation: CS i updates Ni and constructs a bipartite
graph Gi = (Ni , Ci , Li ), where Li represents a set of edges connecting
CSs in Ni and channels in Ci . The weight of edge l = (n, c) ∈ Li
(n ∈ Ni , c ∈ Ci ) is defined as the channel quality of channel c, i.e.,
w(l) = Qc . Let Si denote the set of complete subgraphs of Gi . The
weight of one graph is defined as the weight sum of all edges. Let
s∗i = (Ni∗ , Ci∗ , Li∗ ) denote the complete subgraph in Si with the largest
weight (also termed as MEB [9]). s∗i plays a key role in calculating
the weight of CS i (Wi ). Specifically, Wi is given by
 E
Wi = μ w(l) + (1 − μ) i (2)
∗ E i + j∈N ∗ Ej
l∈Li i
IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 3, JUNE 2019
(a)
(b) (c)
Fig. 1. Example CRSN. (a) 12 CSs {a, b, c, d, e, f , g, h, i, j, k, l} and 5
channels {1, 2, 3, 4, 5}. (b) Bipartite graph of CS d, Gd . (c) MEB of CS d, s∗d .
Algorithm 1 NSAC
1: Initialization: (Ei , Ci ), the set of free CSs F = {1, 2, ..., N}, the set of
clusters CL = φ. %φ denotes the empty set.
2: repeat
3: Update Ni and calculate Wi according to (2) .
4: Target î such that Wî = max Wi ;
i∈F
5: Move CS î and Nî from F to CL, i.e., F = F \ {î} \ N ∗ and CL =
  ∗  î
CL {î} N ;
î
6: until F = φ;
7: Output: CL.
where μ is a balance factor between network stability and remaining
energy.
Fig. 1 employs an example CRSN to illustrate the fundamental
concepts used in the weight calculation of CSs.
Clustering Process: CS i broadcasts Wi and one CS with the largest
weight, say î, is selected as cluster head (CH). The CH î and its
neighbors in N ∗ form one cluster and are marked “clustered.” The
î
remaining CSs update their (Ni , Wi ) and repeat the above clustering
process until all CSs have been marked clustered.
From Algorithm 1, we conclude that NSAC is a fully distributed
protocol since a coordinating center is not required during the
clustering process.
IV. S IMULATIONS
This section simulates the performance of NSAC. We compare
NSAC with CogLEACH [7] and biclique-construction-based cluster-
ing (BCC) [8]. One randomly generated topology is shown in Fig. 2.
The simulation time is divided into equal short periods. Spectrum
states may change across periods. The basic energy consumptions of
CSs in a reclustering period and a stable period are e1 and e2 , respec-
tively. In addition, each CH has to spend extra energy (in proportion
to the cluster size) in managing the cluster. Simulation parameters
are set as follows: N = 100, M = 5, E0 = 5J, ε = 5, μ = 10−4 ,
e1 = 0.01J, e2 = 0.001J, (p1 , p2 , p3 , p4 , p5 ) = (0.8,0.7,0.6,0.5,0.4),
and (Q1 , Q2 , Q3 , Q4 , Q5 ) = (4.1569, 2.5859, 1.7173,
1.1233, 0.7149).
Fig. 3(a) presents clustering intervals (CIs) of clustering protocols.
Obviously, the average and the maximal CIs of NASC are much larger
than those of CogLEACH and BCC, which implies a more stable
CRSN topology of NSAC. Fig. 3(b) shows that the remaining energy
of NASC in the same period is higher than those of CogLEACH
and BCC. We also observe that the network lifetime (defined as the
lifetime of the first dead CS in the CRSN) of NASC is much longer
5865
Fig. 2. Simulated topology; “·” and “” denote CSs and PUs, respectively;
radii of CSs and PUs are 30 m and 70 m, respectively.
(a) (b)
Fig. 3. Performance comparison; NM-network lifetime. (a) CI. (b) Remaining
energy.
than those of CogLEACH and BCC. The advantages of NSAC over
existing methods are attributed to the simultaneous consideration of
spectrum dynamics and energy consumption during the clustering
procedure.
V. C ONCLUSION
A novel NSAC protocol for CRSNs has been proposed. Different
from traditional clustering protocols, this letter has considered
both energy consumption and spectrum dynamics while designing
NSAC.
R EFERENCES
[1] J. Chen et al., “Narrowband Internet of Things: Implementations and
applications,” IEEE Internet Things J., vol. 4, no. 6, pp. 2309–2314,
Dec. 2017.
[2] K. Wang, Y. Wang, Y. Sun, S. Guo, and J. Wu, “Green industrial Internet
of Things architecture: An energy-efficient perspective,” IEEE Commun.
Mag., vol. 54, no. 12, pp. 48–54, Dec. 2016.
[3] J. Wu, S. Guo, J. Li, and D. Zeng, “Big data meet green challenges:
Big data toward green applications,” IEEE Syst. J., vol. 10, no. 3,
pp. 888–900, Sep. 2016.
[4] M. Zheng, L. Chen, W. Liang, H. Yu, and J. Wu, “Energy-efficiency
maximization for cooperative spectrum sensing in cognitive sensor
networks,” IEEE Trans. Green Commun. Netw., vol. 1, no. 1, pp. 29–39,
Mar. 2017.
[5] R. Deng, J. Chen, C. Yuen, P. Cheng, and Y. Sun, “Energy-efficient coop-
erative spectrum sensing by optimal scheduling in sensor-aided cognitive
radio networks,” IEEE Trans. Veh. Technol., vol. 61, no. 2, pp. 716–725,
Feb. 2012.
[6] M. Ozger, F. Alagoz, and O. B. Akan, “Clustering in multi-channel cog-
nitive radio ad hoc and sensor networks,” IEEE Commun. Mag., vol. 56,
no. 4, pp. 156–162, Apr. 2018.
[7] R. M. Eletreby, H. M. Elsayed, and M. M. Khairy, “CogLEACH: A
spectrum aware clustering protocol for cognitive radio sensor networks,”
in Proc. IEEE CROWNCOM, 2014, pp. 179–184.
[8] M. Bradonjić and L. Lazos, “Graph-based criteria for spectrum-aware
clustering in cognitive radio networks,” Ad Hoc Netw., vol. 10, no. 1,
pp. 75–94, Jan. 2012.
[9] D. S. Hochbaum, “Approximating clique and biclique problems,” J.
Algorithms, vol. 29, no. 1, pp. 174–200, Oct. 1998.