Professional Documents
Culture Documents
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
AbstractA novel on-demand cluster-based hybrid routing (SUs), to use licensed bands without interfering the licensed
protocol for cognitive radio ad hoc network with non-uniform users, known as primary users (PUs) [5].
node distribution is proposed in this paper. At first, a novel Cognitive radio networks (CRNs) can be categorized as
spectrum-aware clustering mechanism is introduced. The pro-
posed clustering mechanism divides node into clusters based infrastructure-based CRN and cognitive radio ad hoc networks
on three values: spectrum availability, power level of node and (CRAHNs) [5]. The former CRN has a central network entity
node stability. Therefore, clusters are formed with the highest to maintain the network such as base-stations (BS) in cellular
stability to avoid frequent reclustering. Later, a routing algorithm networks. Instead, the latter does not have any infrastructure
is introduced to minimize the delay while achieving acceptable backbone and each CR user is able to communicate with other
delivery ratio. In this paper, routing is defined as a multi-objective
optimization problem to combine different individual routing CR users via ad hoc connections, [6]. Most of the work in the
metrics to form a global metric. Simulation results show that our field of CRAHN has focused on channel scarcity problem at
proposed routing algorithm can guarantee a lower delay and a the lower layer (PHY, MAC), while routing in CRAHN is
higher packet delivery ratio than conventional routing protocols largely unexplored. Routing in CRAHN is a very important
for cognitive radio ad hoc networks. Due to our routing design task having a great effect on the overall performance of the
specification, such as power consideration and low delay, it can
be suitable to be adopted for IoT applications. network. Routing in CRAHN differs from routing in traditional
ad hoc networks as it has to adapt to dynamic changes of
spectrum due to stochastic behavior of PU and SU. Moreover,
I. I NTRODUCTION routing protocols in CRAHN must deal with heterogeneity of
Due to the expansion of wireless communications, it is resources (available channels and available energy).
almost impossible to cope with allocation of spectrum as it In this paper, we design an efficient clustering mechanism,
is a limited and finite resource. Therefore, it is vital to ensure and on top of that, we propose a robust routing algorithm
the best usage of this limited resource to meet the recent to ensure a certain level of QoS in the network. A novel
radio frequency demand. The traditional spectrum allocation spectrum-aware clustering algorithm is introduced that jointly
methods that are authorized by different regulators are not considers the available spectrum and the power level of the
intelligently handled. It is noticed that the licensed spectrum nodes to form the clusters. The proposed clustering algorithm
remains unoccupied most of the time, [1], where the regulatory makes the network more scalable to PUs activities and
body is unable to allocate the frequency for upcoming new node mobility. The proposed on-demand cluster-based hybrid
wireless devices, [2]. This provisional unused segment in the routing (OCHR), is a hybrid protocol that uses a proactive
licensed spectrum are known as spectrum holes, [3]. Spectrum method for intra-cluster routing and a reactive method for
hole is define as the frequency band that has been allocated to a inter-cluster routing. The path selection in OCHR protocol
licensed user but it is unutilized at a particular time or location. is defined as a multi-objective optimization problem, where
To utilize these unoccupied band, to resolve the spectrum OCHR selects the path with the highest rate and lowest delay.
scarcity for new wireless application and to make spectrum This can be achieved by selecting the route with the lowest
allocation more dynamic, cognitive radio (CR) technology is node density. Using numerical simulations, it is found that
proposed, [4]. This intelligent technology introduces oppor- OCHR outperforms other existing protocols by maintaining
tunistic access of unlicensed users, known as secondary users lesser delay with a higher delivery ratio.
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
rent speed. In the architecture, CR nodes exchange their avail- where Wij is a normalization factor that indicates how well
able channels list (ACLi (t)) according to the spectrum sensing node ij is powerful and static in relation to the other nodes
information. In neighbor discovery phase, every node forms its in the cluster. ij , is a positive value that indicates the
own adjacent list Ni where i Z+ {i = 1, 2, 3, . . . , n}. Af- relationship between the node speed (Eij ) and energy (Vij ).
terwards, based on the exchanged information between nodes, Based on the application requirement, and are design
the cluster formation phase starts that is defined as a maximum parameters for prioritising the speed and energy.
vertex biclique problem, [14]. The goal is while ensuring the To avoid a large CHEV value, we select the log of CHEV
availability of sufficient number of idle channels for intra- as a last selection metric of CHs. Hence, the maximization
cluster communication inside each cluster, accommodate more problem in (1) can be written as follows:
nodes inside each cluster to reduce the control overhead.
Channel-grouping schemes are widely used in ognitive radio ij = max (log (Wij Nij Chij )). (5)
ij
literature for improving the fairness and network stability [15].
At first, every CRi node, i = 1, 2, ..., NSU , creates an This maximization problem can be simply solved using a
undirected bipartite graph Gi (Ai , Bi , Ei ) based on neighbor descending sorting algorithm such as [16]. Therefore, a node
list, Ni and ACLi (t). Graph G (V, E) is called bipartite if that has highest log(CHEV) value among others forms the
vertices set
S V can be split into two disjoint sets A and B cluster and becomes the cluster head (CH). For example, when
where A B = V, such that S all edges in E connect vertices node a and d have 4 shared channels, while node d has a
from A to B. Here, Ai = CRi Ni , and Bi = Ci where, Ci is higher number of neighbors compared to node a under the
the ACLi (t) of CRi . An edge (x, y) exists between vertices x similar power and dynamicity condition, node d will be the
Ai and y Bi if y Ci , i.e., channel y is in the available CH. If the log(CHEV) value of a node CRi is smaller than
channels of CRi . one of its neighbors, CRi joins the neighbour with highest
To select an optimal CHs among all the nodes, a parameter log(CHEV) value, as cluster member (CM). CHs prioritize
named a cluster head election value (CHEV) is defined. In other cluster members based on their value for the reserved
this paper, selection of a CH is formulated as a maximization cluster head (RCH) selection. Once the clusters are formed,
problem, which is defined as follows: The RCH takes over CH responsibilities if current CH moves
out or die, which reduces the possibility of re-clustering. The
ij = max (CHEV ij ), 1 ij CM j , (1) node with the higher log(CHEV) becomes the RCH for the
ij
cluster.
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
A. Path optimization
1
Minimizing the hop-count is one the major consideration F (r) = (DT (r), , HT (r)). (11)
RT (r)
in most of the routing protocols in different areas of wireless
network [17]. However, considering that improvement of dif- To solve the above multi-objective optimization, we use
ferent routing metrics may result in different routing paths, it very common method called weighted sum method [18] for
is necessary to combine different individual routing metrics its simplicity.
to form a global metric to achieve a performance trade-off
n
among different routing metrics. In this paper, as we consider X
M (r) = min wi fi (r), (12)
to improve multiple factors in routing such as delay, data
i=1
rate and hop-count, path optimization is defined as a multi-
objective optimization as follows: where M (r) corresponds to the global objective function F
and r defines the chosen route. n is the number of objective
n
P
min F (x) = (f1 (x) , f2 (x) , ..., fn (x))
T functions and wi > 0, 1 i n. wi = 1 is a weight vector
x (6) i=1
s.t. n 2, fi (x) R+ , defined by each node and reflects the relative importance of
that particular object.
where n is the number of objective functions, F (x) is a vector In CRAHN, the dynamically varying nature of PUs activity
of objective functions fi (x). It is needed to be noted that there may increase the round trip time (RTT). Therefore, RTT
is no single global solution to multi-objective optimization estimation for link level and end-to-end may show abrupt
problems and it is more of a concept than a definition. variation. Thus, this sudden variation may prevent the delay-
Consider the following definitions: based routing protocol to operate correctly. However, higher
(i, j) E = Link from node i to node j; i, j V . data rate assures faster transmission, where values of dqij and
dtij R+ = Transmission delay of link (i, j). dtij are low. Meanwhile, chances of collision are increased in
dbij R+ = Backoff delay of link (i, j). a highly dense network, where increased chance for collision
dqij R+ = Queueing delay of link (i, j). increases dqij . This is because the packet should remain in the
dsij R+ = Switching delay of link (i, j). queue longer till the node finds an available free channel to
R(t)ij R+ = Achievable data rate of link (i, j) at transmit. Also it increases the dbij , as nodes have to perform
time t. backoff procedure more often due to collisions. Therefore, dqij
PT H (s, d) = Path connecting source s to and dbij have direct relationship with network density.
destination d.
dqij , dbij ,
|PT H (s, d)| R+ = Cardinality of PT H (s, d).
d(PT H (s, d)) = Delay of the path PT H (s, d), given 1
dqij , dtij .
by: RT (r)
According to [19], the 1-hop network density is given by:
X
d(PT H (s, d)) = (dtij + dbij + dqij + dsij ). 2
(i,j)PT H (s,d)
N (tr)
(r) = , (13)
(7) A
Using the above definitions, routing for CRAHN may be where N is the total number of nodes in the network, tr is
defined as a multi-objective optimization that finds the route r the transmission range of each node and A is the size of the
from all available routes while simultaneously minimizing the area where the network is deployed.
following objective functions namely: total end-to-end delay However, in this paper, we are trying to choose a route
DT (r), hop count HT (r) and at the meantime maximizing the that passes through the lower dense area as we consider non-
data rate RT (r): uniform distribution. Therefore, we differentiate the density
1- Total end-to-end delay of a route r: in different area of the network and we consider density on
any areas as the number of neighbors that a node has. Higher
DT (r) = d(PT H (s, d)). (8) number of neighbors means higher node density in that area.
2- Hop count between source to destination in route r: Since all traffic flows through CHs, we only consider node
densities around CHs where the number of neighbors for each
CH represents node density . Therefore, total node density
HT (r) = |PT H (s, d)| 1. (9)
in each route r is calculated as follows:
3- Achievable data rate of a route r over HT (r) hop:
2
X NkCM (trk )
C
T (r) = . (14)
1 X A
RT (r) = R(t)ij . (10) kPT H (s,d)
HT (r)
(i,j)PT H (s,d)
where N CM is defined as the number of cluster member.
Therefore, the cost function of a path can be expressed by Data rate of each node is calculated similarly to [20].
the following equation: Defining Rs (t) as the achievable data rate while PU correctly
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
Hop count
20
Rf (t) = (Pon Pd ) log2 (1 + ), (16)
T n(t) + Srp 15
where Pon and Pof f are defined as the probability of a
channel being in occupied state and channel being in idle state, 10
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
2 30 30
1.8 25 25
1
Achievable data rate
Hop count
Hop count
1.6 20 20
1.4 15 15
1.2 10 10
1 5 5
0.01 0.2 0.3 0.4 0.5 0.01 0.2 0.3 0.4 0.5 1 1.2 1.4 1.6 1.8 2
Total node density Total node density Achievable data rate1
(a) (b) (c)
1 1
Fig. 3: Two dimentional projection of Fig. 2: (a) C
T (r) - RT (r) , (b) C
T (r) - HT (r) and (c) RT (r) - HT (r)
the inverse of the achievable data rate RT1(r) and the Z axis delay increases at a higher rate due excessive overhead traffic.
represents the hop count HT (r). Fig. 2 is used to show the Since the TDMA frame size in SCEEM is 300 kb/s, by
impact of the proposed three objective functions to decide the increasing the data rate over 300 kb/s SCEEM shows sharper
route. In Fig. 2, the blue dots represent the different routes increase in delay.
from a source node to the destination node where the red dots The proposed OCHR protocol is also compared with
are the Pareto-optimal solutions. As the performance of the SEARCH and SCEEM protocols in terms of delivery ratio
algorithm is tough to explain from the 3D Pareto-surface, three (Fig. 4(b)). It is shown that the packet delivery ratio decreases
projections of the Pareto-surface onto the corresponding two- for all the three protocols with increasing data rate. Moreover,
dimensional planes are provided in Fig. 3. In this experiment it is also observed that the OCHR can achieve higher delivery
100 different routes are considered from the source node to the ratio compared to the other two protocols. This is due the
destination node. Then, for each route the achievable data rate, consideration of data rate in the path optimization problem
route density and hop-count are calculated. As explained in the where more priority is given to the route that can guarantee
previous section, the Pareto-optimal route, DWA is used to find higher data rate (bandwidth) while the network is not dense.
the best possible routes among all routes. As it can be seen Hence, the proposed OCHR performs better than SEARCH
from Fig. 3, in some of the Pareto-optimal routes may show and SCEEM in terms of delay and delivery ratio (Fig. 4).
lower data rate compared to others, however it is observed that Figure 5 shows the effect of increasing the number of PUs
these routes provide lower density and lesser hop-count at the in the network on the routing protocols while source data rate
same time. In this simulation we give same priority to all the is fixed to 100 kb/s. Fig. 5(a) shows by having more PUs in the
objective functions to obtain the solution. However, depending network the average delay for all the routing protocols increase
on each specific application, high priority might be given to a due to less availability of channels. SCEEM shows the highest
specific objective. For instance, more priority might be given increase in delay. The reason is because the traffic channel
to data rate when the application requires higher bandwidth or that is negotiated between transmitting and receiving nodes
in a delay sensitive application higher priority would be given might not be available due to PUs activity, thus the nodes
to delay. have to go through the negotiation phase again. SEARCH is
The routing efficiency of the proposed OCHR protocol is performing better than SCEEM due to its main mechanism
investigated for end-to-end delay and delivery ratio. Fig. 4 to avoid PUs active regions. However, when the number of
Shows the results against various source data rate and Fig. PUs increases the packet has to route from inefficient route
5 depict the results against increasing the number of PUs which can cause extra delay. OCHR perform slightly better
in the network. The simulation results are compared with than SEARCH due to its clustering scheme in which several
the SEARCH protocol and the SCEEM protocol. Here, we channel are negotiated inside each cluster, thus cluster member
consider the data rate of SU varies from 100 Kb/s to 700 nodes are able to quickly switch between them to avoid extra
Kb/s. Fig. 4(a) presents the comparisons in terms of average delay. However, the inter-cluster routing would impose extra
delay. As it can be seen OCHR shows a significantly lower delay due to more number of PUs activity.
delay compared with SEARCH, meantime, achieving better The effect of increasing number of PUs on protocols deliv-
performance compared to SCEEM. It is due to the high ery ratio is illustrated in Fig. 5(b). As it could be seen from
efficient clustering approaches that SCEEM and OCHR are Fig. 4(b), SCHEEM and OCHR performing slightly better than
using for the routing. At the same time, OCHR can achieve SEARCH when the source data rate was 100 kb/s. However,
better delay than SCEEM due to its adoptive routing algorithm by increasing number of PUs in the network, delivery ratio
that considers few parameters at the same time. It is also of SCEEM drop sharply due to impact of PUs activity on
observed from Fig. 4(a) that average delay in OCHR increases the availability of traffic channel. Meanwhile, SEARCH and
slightly with the increase of the data rate, while for SEARCH OCHR performing better than SCEEM due to their mechanism
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
4 1
SCEE
M SCEEM
SEARCH 0.9 SEARCH
3.5
OCHR OCHR
0.8
3
0.7
Average delay (sec)
2.5
0.6
Delivary ratio
2 0.5
0.4
1.5
0.3
1
0.2
0.5
0.1
0 0
100 200 300 400 500 600 700 100 200 300 400 500 600 700
Source data rate (kb/s) Source data rate (kb/s)
(a) (b)
Fig. 4: Impact of increasing data rate on the performance of protocols in terms of average delay(a) and delivery ratio (b).
3 1
SCEEM SCEEM
SEARCH SEARC
H
OCHR 0.9 OCHR
2.5
0.8
2
Average delay (sec)
Delivary ratio
0.7
1.5
0.6
1
0.5
0.5
0.4
0
16 20 24 28 32 36 40 16 20 24 28 32 36 40
Number of PUs Number of PUs
(a) (b)
Fig. 5: Impact of increasing number of PUs on the performance of protocols in terms of average delay(a) and delivery ratio
(b).
to deal with the PUs activity. As it discussed in Fig. 5(a), intra- a cluster, priority on number of neighbors and bandwidth
cluster routing in OCHR would effect less by increasing the have been given. The proposed OCHR was compared with
number of PUs. However, the impact of increasing the PUs in the other existing routing protocols for CR ad hoc networks
the network would affect the inter-cluster routing and reduce using numerical simulations. It is found that the proposed
the delivery ratio. OCHR outperforms other existing protocols by maintaining
lower delay with higher delivery ratio. In this work, SUs are
VII. C ONCLUSION considered to be fixed and the chosen path is assumed to
be available for complete data transmission. Considering SUs
This paper presented a novel on-demand hybrid routing
mobility and dynamic resource allocation along the path are
protocol for spectrum aware cluster-based network. The novel
remained as our future work.
clustering mechanism presented in this paper considers spec-
trum availability, power level of node and node stability
when forming the clusters. Re-clustering was minimized by ACKNOWLEDGMENT
constructing clusters with the highest stability cluster-heads.
The proposed routing protocol, OCHR, is a hybrid protocol This work is partially supported by National Telecom
that uses a proactive method for intra-cluster routing and a Regulatory Authority (NTRA) Egypt under project title
reactive method for inter-cluster routing. The OCHR routing LTE/WiFi/WiGig internetworking for future 5G cellular net-
protocol was defined as a multi-objective optimization prob- works. We also extend our sincere appreciation to the Dean-
lem, where different individual routing metrics are considered ship of Scientific Research at King Saud University for funding
to form a global metric. To provide the routing weight for this Prolific Research Group (PRG-1436-16).
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2626721, IEEE Access
R EFERENCES
[1] FCC Spectrum Policy Task Force, Report of the spectrum efficiency
working group, 2002.
[2] G. Staple and K. Werbach, The end of spectrum scarcity [spectrum
allocation and utilization], Spectrum, IEEE, vol. 41, no. 3, pp. 4852,
2004.
[3] R. Tandra, S. M. Mishra, and A. Sahai, What is a spectrum hole and
what does it take to recognize one? Proceedings of the IEEE, vol. 97,
no. 5, pp. 824848, 2009.
[4] J. Mitola and G. Q. Maguire Jr, Cognitive radio: making software radios
more personal, Personal Communications, IEEE, vol. 6, no. 4, pp. 13
18, 1999.
[5] I. F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohanty, Next
generation/dynamic spectrum access/cognitive radio wireless networks:
a survey, Computer Networks, vol. 50, no. 13, pp. 21272159, 2006.
[6] I. F. Akyildiz, W.-Y. Lee, and K. R. Chowdhury, Crahns: Cognitive
radio ad hoc networks, Ad Hoc Networks, vol. 7, no. 5, pp. 810836,
2009.
[7] R. M. Eletreby, H. M. Elsayed, and M. M. Khairy, Cogleach: A
spectrum aware clustering protocol for cognitive radio sensor networks,
in Cognitive Radio Oriented Wireless Networks and Communications
(CROWNCOM), 2014 9th International Conference on. IEEE, 2014,
pp. 179184.
[8] M. Ozger and O. B. Akan, Event-driven spectrum-aware clustering
in cognitive radio sensor networks, in INFOCOM, 2013 Proceedings
IEEE. IEEE, 2013, pp. 14831491.
[9] G. Shah, F. Alagoz, E. Fadel, O. B. Akan et al., A spectrum-aware
clustering for efficient multimedia routing in cognitive radio sensor
networks, Vehicular Technology, IEEE Transactions on, vol. 63, no. 7,
pp. 33693380, 2014.
[10] G. Shah, O. B. Akan et al., Cognitive adaptive medium access
control in cognitive radio sensor networks, Vehicular Technology, IEEE
Transactions on, vol. 64, no. 2, pp. 757767, 2015.
[11] V. Esmaeelzadeh, R. Berangi, E. S. Hosseini, and O. B. Akan, Stochas-
tic backlog and delay bounds of generic rate-based aimd congestion
control scheme in cognitive radio sensor networks, Pervasive and
Mobile Computing, 2015.
[12] Q. Zhao, S. Geirhofer, L. Tong, and B. M. Sadler, Opportunistic
spectrum access via periodic channel sensing, Signal Processing, IEEE
Transactions on, vol. 56, no. 2, pp. 785796, 2008.
[13] I. F. Akyildiz, W.-Y. Lee, M. C. Vuran, and S. Mohanty, A survey on
spectrum management in cognitive radio networks, Communications
Magazine, IEEE, vol. 46, no. 4, pp. 4048, 2008.
[14] R. Peeters, The maximum edge biclique problem is np-complete,
Discrete Applied Mathematics, vol. 131, no. 3, pp. 651654, 2003.
[15] J. Zhu, Y. Song, D. Jiang, and H. Song, Multi-armed bandit channel
access scheme with cognitive radio technology in wireless sensor
networks for the internet of things, IEEE Access, vol. 4, pp. 4609
4617, 2016.
[16] D. E. Knuth, The art of computer programming: sorting and searching.
Pearson Education, 1998, vol. 3.
[17] Q. Du, H. Song, Q. Xu, P. Ren, and L. Sun, Interference-controlled d2d
routing aided by knowledge extraction at cellular infrastructure towards
ubiquitous cps, Personal and Ubiquitous Computing, vol. 19, no. 7, pp.
10331043, 2015.
[18] R. T. Marler and J. S. Arora, The weighted sum method for multi-
objective optimization: new insights, Structural and multidisciplinary
optimization, vol. 41, no. 6, pp. 853862, 2010.
[19] N. Bulusu, D. Estrin, L. Girod, and J. Heidemann, Scalable coordina-
tion for wireless sensor networks: self-configuring localization systems,
in International Symposium on Communication Theory and Applications
(ISCTA 2001), Ambleside, UK, 2001.
[20] G. Shah, O. B. Akan et al., Csma-based bandwidth estimation for
cognitive radio sensor networks, in New Technologies, Mobility and
Security (NTMS), 2012 5th International Conference on. IEEE, 2012,
pp. 15.
[21] Y. Jin, M. Olhofer, and B. Sendhoff, Evolutionary dynamic weighted
aggregation for multiobjective optimization: Why does it work and how,
in Genetic and evolutionary computation conference, 2001, pp. 1042
1049.
[22] Y. Jin, T. Okabe, and B. Sendho, Adapting weighted aggregation
for multiobjective evolution strategies, in Evolutionary multi-criterion
optimization. Springer, 2001, pp. 96110.
[23] K. R. Chowdhury and M. D. Felice, Search: A routing protocol for
mobile cognitive radio ad-hoc networks, Computer Communications,
vol. 32, no. 18, pp. 19831997, 2009.
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.