High Throughput Transmission Quality Aware Broadcast Routing in Cognitive Radio Networks

Noname manuscript No.(will be inserted by the editor)

High-Throughput Transmission-Quality-Aware BroadcastRouting in Cognitive Radio Networks

Ejaz Ahmed, Junaid Qadir, and Adeel Baig

the date of receipt and acceptance should be inserted later

Abstract Cognitive radio is an enabling technology of dynamic spectrum ac-cess (DSA) networking. In DSA, unlicensed secondary users can coexist withprimary licensed users and can share the radio spectrum opportunistically.Broadcasting is an important networking primitive that is useful for manyCRN applications such as control information dissemination, warning noti-fication, etc. Unfortunately, the sporadic channels availability degrades theperformance of broadcast routing. The quality of a broadcast transmissionon a particular channel depends on the channel quality of all the receiversfor the same transmitter. Current broadcast routing protocols lack transmis-sion quality awareness. In this paper, we develop a transmission quality-awarebroadcasting framework, comprising algorithm for transmission quality-awarebroadcast routing in multi-radio dynamic-spectrum-access CRNs, and formu-late a transmission quality metric wherein we consider a receiver-centric viewrather than a transmission-centric view. We perform a detailed simulationperformance evaluation of our proposed framework using OMNeT++. Theproposed broadcast routing algorithm is validated by comparing results withstate-of-the-art routing algorithms. Analysis of the results shows average per-formance gains of approximately 40 percent in throughput and packet deliveryratio.

Keywords Cognitive Radio Networks · Transmission Quality · BroadcastRouting Metric · Distributed Broadcast Routing

Ejaz Ahmed (corresponding author) is with Faculty of Computer Science and Infor-mation Technology, University of Malaya, Kuala Lumpur, Malaysia. Junaid Qadir andAdeel Baig are with SEECS, National University of Sciences and Technology (NUST),Islamabad, Pakistan. E-mail: [email protected], [email protected],[email protected]

Address(es) of author(s) should be given

2 Ejaz Ahmed, Junaid Qadir, and Adeel Baig

1 Introduction

With the advancement in networking technologies, several new distributedcomputing paradigms are emerging such as Mobile Cloud Computing [1–3] andVehicular Cloud Computing [4]. These emerging computing paradigms haveenabled a number of new broadcast applications such as publicity systemsand cloud-based warning notification systems [5]. However, performance ofsuch applications is based on routing in the wireless networks. The problem ofrouting is a heavily researched area in wired networks [6–8] as well as in severalwireless networks, such as mobile networks [9–12], wireless sensor networks [13–16], delay tolerant networks [17–19], and in cognitive radio networks (CRNs)[20–22]. The problem of broadcast routing in CRNs is particularly challengingdue to the complexities involved in managing: (i) varying number of availablechannels [23]; (ii) intermittent availability of channels [24, 25]; and (iii) theredundancy of transmissions [26].

Broadcasting performance in CRNs can benefit significantly if the broad-cast routing framework can exploit the available diversities in multi-radioCRNs, such as the radio-diversity (the ready availability of the options of mul-tiple radio interfaces on CRN nodes); the channel-diversity (the availability ofmultiple radio channels to which a radio belonging to a CRN node can tuneto); and the rate-diversity (the ability of a modern wireless network interfacecard to tune to any of a number of different transmission rates). Exploitingmulti-radio interfaces and multi-channel in routing increases the throughput byallowing multiple transmissions in parallel if the interfaces on a node are tunedon orthogonal channels [27–29]. The exploitation of multi-rates, on the otherhand, helps in making appropriate range and throughput tradeoffs for highperformance broadcasting performance [30]. In addition to exploiting these di-versities, a broadcasting framework should also exploit the “wireless broadcastadvantage” (WBA) [31] due to which a transmission by a node can be receivedby all neighboring nodes that lie within its communication range (assumingthe use of omnidirectional antennas).

Due to the dynamic channel conditions that characterize CRNs, the chan-nels available to a node have significant temporal and spatial variations. Thechannel quality also varies for different available channels of a node. Anybroadcast framework that hopes to be practically efficient should incorpo-rate the variable channel quality into its design. A key insight here is that asingle broadcast transmission can have multiple receivers, with each receiverpotentially perceiving different channel quality (for the same transmission).The transmission-quality of a broadcast transmission on a channel, therefore,should consider the channel quality as a function of the channel quality per-ceived by individual receivers. While earlier research efforts have modeled thequality of transmissions in the context of unicast communication (e.g., [32]),transmission quality has not been considered for broadcast transmission yetin literature.

In our previous work on broadcasting in WMNs [30] [33] [34], we hadfocused on minimum latency broadcasting—defined as ‘the maximum delaybetween the transmission of a packet by the source node and its eventual

Title Suppressed Due to Excessive Length 3

reception by all receivers’. These works focused on the analysis of broadcastlatency for a single packet, not taking into account intra-flow interference whichcan be better measured by throughput. In this work, our metric of choice isthroughput and packet delivery ratio (PDR), and our focus is exclusively onthe design of high throughput transmission-quality-aware broadcast routingin CRNs. We incorporate transmission quality awareness in our frameworkand assess the performance of our algorithms in terms of metrics of PDR andthroughput.

Contributions of this paper. This paper proposes a novel transmission-quality-aware distributed framework for broadcast routing in CRNs. To caterto the dynamic nature of CRNs, we have built adaptive features into our broad-casting framework to ensure that appropriate action is taken in response tochange in transmission quality. Our aim is creation of high-throughput broad-casting in CRNs, and therefore, we have only considered throughput and PDRand not delay as performance metric. To the best of our knowledge, this isthe first work to incorporate a receiver-based transmission quality awarenessmetric into the problem of broadcasting. Our contributions in the paper in-clude: (a) a novel formulation of transmission quality metric for CRNs whichincorporates all the receivers’ channel quality of broadcast transmission and(b) design and analysis of an adaptive transmission-quality-aware broadcastrouting algorithm in CRNs which employs the metric. The proposed frame-work is evaluated in the simulation environment by using OMNeT++. Theanalysis of the results shows up to 40 percent improvement in throughput andpacket delivery ratio for the proposed framework while comparing with currentbroadcast routing algorithms.

Organization of this paper. The rest of this paper is organized as follows.We present the related work in Section 2. In Section 3, we illustrate the packetloss ratio (PLR) information exchange mechanism that we have used in ourproposed framework. In Section 4, we present details of our proposed broadcastrouting framework. In Section 5, we discuss the network model that we haveconsidered. In Section 6, we present the performance evaluation of the pro-posed routing framework after first discussing our simulation model. Finally,the paper is concluded in Section 7.

2 Related Work

Broadcasting in wireless networks is a fundamentally different problem tothe broadcasting problem in wired networks due to the wireless medium’scharacteristic “broadcast advantage” [31]. The metrics typically used to studybroadcasting in wireless networks are energy consumption [31], the number oftransmissions [35], broadcast latency [30] [33] [34], or the overhead in route dis-covery and management. In our work, we have chosen the metric of throughputto incorporate both intra-flow and inter-flow interference in our performanceevaluation framework.


In our previous work, we have proposed broadcast latency minimizationheuristics for multi-radio wireless networks including the centralized “par-allelized approximately-shortest multi-radio WCDS tree” (PAMT) heuristic[34] and the distributed “multi-radio distributed tree” (MRDT) heuristic [36].These heuristics also exploited the rate, radio, and channel diversities, in ad-dition to leveraging WBA—like the algorithms in the paper. These heuristicsare proposed for WMNs which have relatively static conditions. The workpresented in this paper differs from these previous works in that our proposedalgorithms are: (i) designed for CRNs which are highly dynamic; and ii) adap-tive to channel conditions by incorporating transmission quality awareness.

In another broadcasting work for CRNs, Kondareddy et al. [37] introduce acontrol information dissemination mechanism in CRNs in absence of commoncontrol channel. They formulate this problem graph-theoretically and addressit by finding minimal neighbor graphs. In their proposed solution, initially anessential channel set covering all 1 -hop neighbors of a node is found. The es-sential channel set is then used to disseminate the control information, therebyminimizing the overhead of complete broadcast. The solution is similar to ourproposed transmission-quality-aware broadcast routing framework, in the dis-semination of broadcast traffic on only the essential channel set covering allthe one hop neighbors. Rehmani et al. [38] extended the work of Kondareddyet al. [37] and proposed a channel selection for CRN data dissemination. Theproposed channel selection strategy considered both the number of secondaryneighboring nodes on a particular channel and the PU activity on that channel.Our proposed solution differs from the solutions proposed in [37] [38] in thatwe incorporate 2 -hop neighbors information to find the essential channel setand minimal relay nodes set among 1 -hop neighbors to reach all 2 -hop neigh-bors, and that our work focuses on incorporating the transmission-quality intorouting design.

Borges et al. [39] present a taxonomy of routing metrics for WMNs. Themajority of surveyed routing metrics utilize parameters related to quality oflink such as traffic load and interference. Some of the state-of-the-art researchworks integrate these parameters to devise a hybrid metric. These routingmetrics are mostly designed for unicast routing algorithms, which are notapplicable to broadcast routing problem.

Apart from the work on broadcasting, the work done on multicasting isalso related to our work—since broadcasting is a special case of multicasting inwhich the multicast group contains all the network nodes. A multicast routingmetric, Expected Multicast Transmission Time (EMTT), is proposed in [40] formulti-rate multi-hop WMNs to increase the multicast transmission reliability.The routing metric incorporates the MAC-layer rate diversity, retransmission-based reliability, link quality awareness, and wireless broadcast advantage. Arecent survey on the various algorithms, techniques and protocols used formulticasting in CRNs is presented in [41].


3 Packet loss ratio and PU activity ratio information exchange

Before, we explain our proposed transmission quality metric and broadcastrouting framework, we illustrate the PLR and PU activity information ex-change procedure for broadcast. Each node periodically broadcasts PROBEpackets on each of its channel after every 3 seconds. On the receiver side, eachreceiver computes the PLR by considering total packets received and totalpackets sent during interval of every 30 seconds.

In the example topology illustrated in Figure 1, each receiver computesthe PLR for all assigned channels by considering the node ‘a’ as a transmitternode and informs the node ‘a’ about the PLR for all the channels by sendingInformationExchange message. The message comprises of source ID, targetID, number of common channels between two nodes, channel ID of each chan-nel along with corresponding PLR value, and PU activity ratio. The sourceID field is an ID of receiver node which is source of a InformationExchangemessage, the target ID field is an ID of a transmitter node for which the PLRvalue is computed, number of common channels field that tells how many en-tries the message can have, channel ID field identifies the channel for whichPLR value is computed, PLR value field contains the PLR for last 30 seconds,and expected PU activity ratio value for last 30 seconds.

Fig. 1 Packet Loss Ratio and PU ON and OFF Activity Ratio of Nodes on DifferentChannels

For explanation, we consider the node ‘e’ among the receivers in the exam-ple illustrated in Figure 1 for exchange of information. Node ‘e’ has its averagePLR values 0.099, 0.06 and 0.3 for channel 1, 2 and 3 respectively. The PUactivity ratios for channel 1, 2, and 3 are 0.5, 0.2, and 0.4 respectively. It ex-changes the message shown in Figure 2. The message is exchanged in every 30seconds to update the receivers’ PLR and PU activity ratio at the transmitter.

Similar to node ‘e’ each receiver node exchanges the message containingPLR value and PU activity ratio for each channel common with node ‘a’, thennode ‘a’ computes the transmission quality based on the updated availablePLR values and PU activity ratio for all receivers. The process is repeatedevery 30 seconds to update the transmission quality value.


Fig. 2 Format for Information Exchange Message

4 Transmission-quality-aware broadcast routing

We propose a distributed transmission-quality-aware broadcast routing mech-anism that uses the PLR and PU activity ratio on all receivers of a particulartransmission. The routing is based on the transmission quality that is first es-timated, then the estimated transmission quality is enhanced by exporting thebottlenecked receiver either to its own another interface or to another neighbor-ing node. Then, based on the transmission-quality metric a broadcast routingtree is constructed. As CRNs are highly dynamic networks, the path on therouting tree can be broken that need to be repaired to sustain the communi-cation among the nodes. Our proposed broadcast routing framework supportsthe route recovery mechanism to repair the tree. The whole transmission-quality-aware broadcast routing consists of the following four steps.

1. Estimation of transmission quality metric2. Enhancement of transmitted quality

(a) Intra-node enhancement(b) Inter-nodes enhancement

3. Construction of broadcast routing tree4. Recovery of broadcast tree

4.1 Modeling and Estimation of Multicast/Broadcast Transmission QualityMetric

The multicast/broadcast transmission quality metric for estimating the qualityof transmission on a particular channel is modeled by including the receptionquality measured by multiple potential receivers of that transmission. In multi-cast/broadcast, multiple receivers can receive same transmitted packet; hence,transmission quality includes the perceived quality of all the receivers of thatparticular transmission.


The transmission quality metric Qt(n,R, c, t) can be mathematically for-mulated as follows:

Qt(n,R, c, t) = ∀j ∈ R

ln−1

1∏j∈R ln−1(PLR(n,c,j,t) )

×1

ln−1max|[N1]|j=1

E[T ]ON,(j,c,t)E[T ]ON,(j,c,t)+E[T ]OFF,(j,c,t)

(1)

Qt(n,R, c, t) can be read as transmission quality of a node ‘n’ with set ofreceivers ‘R’ on channel ‘c’ at time instance ‘t’ . In equation 1, [N1] representsset of 1 -hop neighbors including the transmitter node. The nodes estimatethe PU channel activity after every 1 second and exchange the expected PUactivity ratio with other nodes using the InformationExchange message.The estimation of the PU is based on the formula stated in equation (9) ofMubashir et al. work [38]. The values of mis-detection and false alarm arekept as 0.1. The values of PLR are less than one so when we directly multiplysmall values of PLR, the resultant value become smaller. Hence, we have touse a function ln−1 to accentuate the value of PLR and the result. In order tobalance the weights of PLR and PU activity ratio, we also have to use ln−1

with PU activity ratio. The E[T ]ON,(i,c,t) and E[T ]OFF,(i,c,t) are the expectedON and OFF times estimated by node i on channel c at time t. The PLRon a particular channel between two nodes can be computed in two ways,actively and passively. The active approach mandates the active exchange ofperiodic PROBE messages between the nodes. Whereas, the passive approachlet the measurement of the transmission quality without exchange of PROBEmessages. The estimation of transmission quality becomes quickly outdatedin passive approach when the data traffic is less. We have used the activeapproach in our simulation, where the PROBE messages are exchanged withthe interval of 3 seconds.

4.1.1 Illustration of Concept by an Example

To illustrate the computation process of transmission quality, we have used asimple network topology that consists of five nodes. For this example, we takethe following assumptions: (a) each node has three interfaces, (b) each interfaceis tuned on the corresponding channel (for example interface one is tuned tothe channel one), and (c) the channels are already assigned to the interfaces inthe network. The transmitter node ‘a’ finds the transmission quality for all itschannels considering the PLR and PU activity ratio on all receiver nodes. ThePLR and PU activity ratio for available channel is given in Figure 1 (a), Figure1 (b), and Figure 1 (c). The PLRs and their corresponding accentuated valuesare given in Table 1 for each node and channel. The transmission quality onnode ‘a’ in Figure 1 is computed as illustrated in Figure 3.


Table 1: PLR inverse natural log and PU activity ratio values for channels ofeach receiver

Channel #‘1’Receivers ‘a’ ‘b’ ‘c’ ‘d’ ‘e’

PLR 0.200 0.150 0.100 0.099ln−1 1.220 1.160 1.110 1.094

E[T ][ON ],(i,c,t) 0.2 0.4 0.3 0.4 0.5E[T ][OFF ],(i,c,t) 0.8 0.6 0.7 0.6 0.5

Channel # ‘2’Receivers ‘a’ ‘b’ ‘c’ ‘d’ ‘e’

PLR 0.100 0.080 0.070 0.060ln−1 1.110 1.080 1.070 1.060


Channel # ‘3’Receivers ‘a’ ‘b’ ‘c’ ‘d’ ‘e’

PLR 0.250 0.150 0.200 0.300ln−1 1.280 1.100 1.220 1.350


Qt(a, 1, b, c, d, e, t) = ln−1

(1

1.220 + 1.160 + 1.110 + 1.094 ×1

ln−1max(

0.20.2+0.8

, 0.40.4+0.6

, 0.30.3+0.7

, 0.40.4+0.6

, 0.50.5+0.5

))

= ln−1(

14.58 ×

11.64

)= ln−1

(1

7.51

)= ln−1 (0.133) = 1.14

Qt(a, 2, b, c, d, e, t) = ln−1

(1

1.110 + 1.080 + 1.070+ 1.060 ×1

ln−1max(

0.40.4+0.6

, 0.20.2+0.8

, 0.30.3+0.7

, 0.10.1+0.9

, 0.20.2+0.8

))

= ln−1(

14.32 ×

11.49

)= ln−1

(1

6.44

)= ln−1 (0.155) = 1.17

Qt(a, 3, b, c, d, e, t) = ln−1

(1

1.280 + 1.100 + 1.220 + 1.350 ×1

ln−1max(

0.30.3+0.7

, 0.30.3+0.7

, 0.30.3+0.7

, 0.20.2+0.8

, 0.40.4+0.6

))

= ln−1(

14.95 ×

11.49

)= ln−1

(1

7.38

)= ln−1 (0.14) = 1.15

Fig. 3 Transmission quality metric computation on node ‘a’ for channel 1, 2, 3

4.2 Enhancement of transmission quality

For enhancement of transmission quality phase, we consider that each interfaceof a node has bin containing set of neighboring nodes. In this phase, thebottlenecked receiver node is moved from bini to binj if and only if productof packet loss ratio and PU activity ratio increases by this move. The biniand binj have set of neighbors covered on channels tuned by radio interface ‘i ’and ‘j ’, respectively. The transmission quality enhancement is followed from


the observations captured in different simulation scenarios that the estimatedtransmission quality on a channel can be mostly impacted by one (or a few)bottlenecked receiver nodes who have high PU activity ratio or poor receptionquality.

In such scenarios, the transmission quality is required to be enhanced bymoving these bottlenecked node(s) to either other radio interface of the samenode (which is referred to as ‘intra-node’ enhancement) or interface of anotherneighboring receiver node (which we refer to as an ‘Inter-nodes’ enhancement).This phenomenon forms the basic intuition for the enhancement of transmis-sion quality.

4.2.1 Intra-node Enhancement

A receiver ‘r’ of a node ‘n’ can be exported from its one interface ‘k’ to itsown another interface ‘l’ when the following conditions occur:

– The receiver ‘r’ can also be covered on interface ‘l’– Previous accumulative product of PLR and PU activity ratio is greater

than that of new settings– New accumulative product of PLR and PU activity ratio is least among

all available export options

∑k∈R1

(PLR(n, c1, k, t)×max(

E[T ]ON,(n,c1,t)

E[T ]ON,(n,c1,t) + E[T ]OFF,(n,c1,t)

,E[T ]ON,(k,c1,t)

E[T ]ON,(k,c1,t) + E[T ]OFF,(k,c1,t)

)

)+

∑l∈R2

(PLR(n, c2, l, t)×max(

E[T ]ON,(n,c2,t)


,E[T ]ON,(l,c2,t)

E[T ]ON,(l,c2,t) + E[T ]OFF,(l,c2,t)

)

)

>∑

k∈R1/r

(PLR(n, c1, k, t)×max(

E[T ]ON,(n,c1,t)


,E[T ]ON,(k,c1,t)


)

)+

∑l∈R2

⋃r

(PLR(n, c2, l, t)×max(

E[T ]ON,(n,c2,t)


,E[T ]ON,(l,c2,t)


)

)(2)

The R1 and R2 represent set of receivers on channel c1 and c2 respectively.Figure 4 shows the estimated transmission quality intra-node enhancementphase employed by the transmission-quality-aware broadcast routing.

4.2.2 Inter-node Enhancement

A receiver ‘r’ of a node ‘n’ can be exported from its interface ‘k’ to anotherneighboring node’s interface ‘l’ when following conditions occur:

– The receiver ‘r’ can also be covered on interface ‘l’ of another neighboringnode ‘r2’

– Previous accumulative product of PLR and PU activity ratio is more thanthat of new settings


Fig. 4 Estimated Transmission Quality Intra-node Enhancement Example

– New accumulative product of PLR and PU activity ratio is least amongall available export options

∑k∈R1

(PLR(n, c1, k, t)×max(E[T ]ON,(n,c1,t)


,E[T ]ON,(k,c1,t)


)) >

∑l∈R1/r

(PLR(n, c1, l, t)×max(E[T ]ON,(n,c1,t)


,E[T ]ON,(l,c1,t)


))

+(1− (1− (PLR(n, c1, r2)×max(E[T ]ON,(n,c1,t)


,E[T ]ON,(r2,c1,t)

E[T ]ON,(r2,c1,t) + E[T ]OFF,(r2,c1,t)

)))

×(1− (PLR(r2, c, r)×max(E[T ]ON,(r2,c1,t)

E[T ]ON,(r2,c1,t) + E[T ]OFF,(r2,c1,t)

,E[T ]ON,(r,c1,t)

E[T ]ON,(r,c1,t) + E[T ]OFF,(r,c1,t)

))))

where r2 ∈ R1/r(3)

Figure 5 shows the estimated transmission quality inter-node enhancementphase employed by the transmission-quality-aware broadcast routing.

4.3 Construction of Broadcast Tree

The tree construction phase is the only source independent phase in our pro-posed broadcast routing framework. In this phase, each node selects its rate,channel using 1 -hop and 2 -hop neighbors information in a distributed manner.The selection of rate and channel is based on better transmission quality andcoverage of maximum number of nodes with minimum delay. In Procedure 1:TransInformation at line 6 and line 7, rate and channel, which gives maximumfunction value, are selected. On line 12 transmission related parameters aresaved in the header field ti(j). The block of code from line 5-15 will be executeduntil all 1 -hop neighbors of node ‘n’ are covered. The block of code line 17-27determines 1 -hop neighbors that will become forwarders to cover the 2 -hop


Fig. 5 Estimated Transmission Quality Inter-node Enhancement Example

neighbors. The flags for the selected forwarder nodes have been set on line 23in the header field. In Procedure 2: SendRREQ, the sender node creates andsends the RREQ message that contains the source node ID, sequence number,node label, sender ID, and transmission information. On reception of RREQmessage, receiver nodes process the RREQ message and identify whether theircurrent label is greater than newer label then they tune their radio interfaceson channel and rate mentioned in the header field.

Table 2 shows the notations and their corresponding definitions used intransmission-quality-aware broadcast routing tree algorithm.

Algorithm 1: Tree Construction Algorithm for Transmission Quality-Based Broadcast Routing

Input: n,A,AN1, N1, N2, Qt(A), QtN(A), Ω,L

1 if a RREQmessageisreceivedonn then2 [Sibling(n), P (n), ζ, ω] = processmsgRREQ(n,RREQ)3 end4 Compute Information of Trasnsmission :5 ti = TransInformation(n,N1, N2, Sibling(n), Qt(A), QtN(A), A,AN1 , Ω, L)6 if n is acting as a source node then7 seqno← 0 for first time and incrementit for new RREQmessages8 Broadcast RREQ using SendRREQ(n, seqno, ti)

9 end10 if n is acting as a forwardernode then11 Broadcast RREQ by executing SendRREQ(n, ti)12 end

Output: P (n), ζ, ω

Each node implements two key functionalities for tree construction: (a) thenode selects its interface that gives the best transmission quality and coversmaximum number of nodes, and (b) the node also selects the forwarder for itstransmission as a next-hop node. Initially, a RREQ is transmitted by source


Table 2: Notations and Their Corresponding DefinitionsSymbol Descriptionn The node on which the algorithm runsSibling(n) Set of sibling of node ‘n’N1 Set of 1-hop neighboursN2 Set of 2-hop neighboursA Set of channels tuned on node ‘n’AN1 Set of channels assigned on node ‘n’ ’s 1-hop neighboursΩ Set of rates that are supported on node ‘n’Qt(A) A vector of transmission quality of channels assigned to a nodeQtN(A) Transmission quality matrix of assigned channels of neighboursL Latencies SetFN1 Set of forwarder nodes for 1-hop neighbouring nodesRREQ Route request broadcast messageRREQ Previously transmitted route request message from same

sourceP (n) Node n’s parent nodeI(n) Node n’s Interface for making link with parent nodeω Set of latencies corresponding to rate used for making link with

parent nodeζ Channel tuned on link with parent nodetr Transmission information of neighbours [Node ID, Channel,

Rate, flag]flag flag represents a forwarder node, flag = 1 mean forwarder/ Difference operator of sets

node on its minimum number of interfaces that can reach all of its neighbors.The message RREQ contains the following information: source ID, sequencenumber, sender ID, neighbor ID with channel and rate on which the neighborcan be covered, and set of forwarders. When the intermediate node receives theRREQmessage, it looks for its own ID into set of forwarders in RREQmessage.If it finds its own ID in set of forwarder list then the intermediate node selectsthe forwarder nodes, channel, and rate to relay the data further. Thereafter,information containing the forwarder ID, channel, and rate is appended toreceived RREQ message that is further sent on minimum number of interfacesthat can cover all the neighboring nodes. The intermediate node does notforward the RREQ message if the node has already sent it with high or samelabel value. The nodes compute its own label value by adding the latency ofthe selected link in sender node’s label. The value of label for source node isset as zero.

4.3.1 Example illustration

In the example illustrated in Figure 6, ‘a’ is taken as the source for the broad-cast. When an application running on a node ‘a’ generates data, the nodeinitiates the proposed transmission quality aware broadcast routing. Initially,


Procedure 1: TransInformation()Input: [n,N1, N2, Sibling(n), Qt(A), QtN(A), A,AN1

, Ω,L]1 if n is acting as a source then2 Sibling(n)← ∅3 end4 Z ← ∅5 while N1/Z 6= ∅ do6 f(n, c, r)← arg max∀c∈A/C,∀r∈Ωf(n, c, r)

7 f(n, c, r)← Qt(c) × |N(n,c,r)/(Sibling(n)∪Z)|l

8 where l is latencycorresponding to rate ‘r′

9 C = C + c where ‘ +′ is aconcatenationoperator10 R← R + r11 Z ← Z ∪N(n, c, r)12 ti(j)← [Covered Node ID, r, c, flag]13 j ← j + 1 for each covered node14 where Covered Node ∈ (N(n, c, r)/(Sibling(n) ∪ Z))15 end16 X ← N2,V← ∅17 i← 1 while N2/X 6= N2 do18 Y ← N(n,C(i),R(i))19 f(y, c, r)← arg max∀y∈Y,∀c∈Ay,∀r∈Ωf(y, c, r)

20 f(y, c, r)← Qty(c)× |N(y,c,r)/(N1∪V)|l

21 X ← X/(N(y, c, r)/(N1 ∪V))22 V← V ∪N(y, c, r)23 if ∃ ID|ID == y&ID ∈ ti then24 set flag in ti25 end26 i← i+ 1

27 endOutput: [ti]

Procedure 2: SendRREQ()Input: n, ti, seqno

1 if n is acting as source then2 RREQ.sourceID ← n3 RREQ.sequenceNumber ← seqno4 RREQ.label← 0

5 end6 RREQ.senderID ← n7 RREQ.txinfo← ti8 Send RREQ on all radio interfaces otherthan from9 which RREQ has received

the node ‘a’ determines its channel and rate by executing the procedure 1:TransInformation as shown in Table 3. We have defined rate latency scale asfollows: 1 Mbps = 1 unit delay and 11 Mbps = 0.09 unit delay. In this ex-ample, a combination of 11 Mbps data rate and channel ‘1’ is chosen sincethe function returns maximum value for the chosen combination as presentedin Table 3. To cover entire set of 1 -hop neighbors, the node ‘a’ again runs


Procedure 3: processmsgRREQ()

Input: n,RREQ,RREQ1 if RREQ.sourceID = RREQ.sourceID then2 if RREQ.sequenceNumber = RREQ.sequenceNumber then3 if RREQ.label + l < label(n) then4 l is the corresponding latency to rate RREQ.ti.n.r

5 if Qtc(I(n))−Qtc(I(n)) > η then6 Send Leave messagetoP (n)7 end8 ζ ← RREQ.ti.channel

9 ω ← RREQ.l10 P (n)← RREQ.senderID

11 label(n)← RREQ.label + l12 Sibling(n)← RREQ.neighbors

13 end14 else if RREQ.label + l > label(n) then15 Send ALREADY COV ERED messagetosender16 end17 end18 end19 return P (n), Sibling(n), ζ, ω

Fig. 6 Network topology for transmission-quality-aware broadcast tree construction


Table 3: Rate and channel selection on node ‘a’

iteration = 0 Covered Nodes= C Qt(c) l Uncov. neighbors f(n,c,r) = Qt(c) ∗ N

l

1 0.9 0.09 c 1 =0.9* 10.09=10

1 0.9 1 c 1 =0.9* 11=0.9

2 0.85 0.09 c 1 =0.85* 10.09=9.44

2 0.96 1 b, c 2 = 0.96* 21=1.92

iteration = 1 Covered Nodes= cC Qt(c) l Uncov. neighbors f(n,c,r) =Qt(c) ∗ N

l

1 0.9 0.09 0 =0.9 * 00.09= 0

1 0.9 1 0 =0.9 * 01=0

2 0.85 0.09 0 = 0.85 * 00.09 =0

2 0.96 1 b 1 = 0.96 * 11 =0.96

iteration = 0 Covered Nodes= b, cC Qt(c) l Uncov. neighbors f(n,c,r) =Qt(c) ∗ N

l

1 0.85 0.09 e,d 2 = 0.85× 20.09 = 18.88

1 0.85 1 e,d 2 = 0.85× 21 = 1.7

2 0.96 0.09 d 1 =0.96× 10.09 = 10.67

2 0.96 1 d 1 =0.96× 11 = 0.96

Covered Nodes= b, c, d, e

the algorithm. Now, combination of data rate of 11 Mbps and channel ‘2’ isselected based on the highest values of 0.96.

As the node ‘a’ has already selected its rates and channels for all of its1 -hop neighbors, now node ‘a’ chooses its set of forwarder nodes among 1 -hopneighbors. The node a chooses the node c as one of the forwarder nodes becausethe node c can cover maximum number of 2 -hop neighbors. After finding theforwarder set that covers all 2 -hop neighbors, node a sends RREQ messageby executing the procedure 2: SendRREQ. The RREQ message contains thechannel, rate and forwarders information for 1-hop neighbors. When the 1 -hopneighbors ‘b’ ‘c’ receive the RREQ message, they process the RREQ messageby running the procedure 3: processmsgRREQ. The node ‘b’ examines themessage and finds that node ‘a’ has not declared the node ‘b’ as forwarder, sothe node ‘b’ does not choose its channel and rate for forwarding the packets.However, the node ‘c’ finds itself in the set of forwarder nodes, therefore itruns the procedure 1: TransInformation. Afterwards, the node ‘c’ choosesrates and channels for its uncovered neighboring nodes. The node ‘c’ choosesrate 11 Mbps and channel ‘1’ for its neighboring nodes ‘d’ and ‘e’ as shown inTable 3.


4.4 Route Recovery

The wireless links along the routes in CRNs are highly dynamic because ofgeneral interference and PU’s activity. Therefore, it is vital to recover theroutes by finding alternative paths if the previously established routes areno longer available. The mathematical representation of our route recoveryalgorithm is presented in procedure 4: RouteRecovery. The route recoveryalgorithm is executed whenever a PU activity is detected by a secondary useron an active link. To cope with such scenarios, the secondary nodes redirectthe outgoing traffic to other interfaces or nodes. The route break problem canbe handled by following methods:

Procedure 4: RouteRecovery()Input: n,N1, N2, FN1

, Sibling,A,AN1, Qt(A), QtN(A), Ω,L

1 if n gets CHANNEL_UNAV AILABLE messagefrom ‘d′ ; d ∈ N2 then2 if d /∈ N(m, c, r)∀m∈FN1

,∀c∈A,∀r∈Ω then3 f(m, c, r)← arg max∀m∈N1/FN1

,∀c∈AN1(m),∀r∈Ωf(m, c, r)

4 f(m, c, r)← Qt(c)×|N(m,c,r)/(Sibling(m)∪∀v∈FN1

N(v,c,r))|

lwhere d ∈ N(m, c, r) where l ∈ L

5 Create RRECOV ERED message6 RRECOV ERED.destinationID ← d7 RRECOV ERED.forwarderID ← m

8 RRECOV ERED.label← ( 1Qt(c)

× 1|N(m,c,r)/Sibling(m)| ) + label(m)

9 RRECOV ERED.neighbors← N(m, c, r)/(Sibling(m)⋃

∀v∈FN1

N(v, c, r))

10 RRECOV ERED.Interface← i11 radiointerface on which channelc is tuned12 Send RRECOV ERED(m, c, r)

13 end14 else if n receives CHANNEL_UNAV AILABLE from node ‘d′

15 where d ∈ N1 then16 f(n, c, r)← arg max∀c∈A,∀r∈Ωf(n, c, r)

17 f(n, c, r)← Qt(c)× |N(n,c,r)/Sibling(n)|l

18 whered ∈ N(n, c, r)19 where l is latencycorresponding to r

20 end21 end22 else if MAC gets CHANNEL_UNAV AILABLE then23 Recover entire set of neighbors covered by previously available channels24 end25 if n gets RRECOV ERED(n, c, r) then26 if RRECOV ERED.forwarderID = n then27 forward RRECOV ERED28 end29 else if RRECOV ERED.destinationID = n then30 Sibling(n)← RRECOV ERED.neighbors/n31 P (n)← RRECOV ERED.forwarderID32 label(n)← RRECOV ERED.label

33 end34 end


1. Covering the affected nodes by the another interface of the same node2. Recover the affected nodes through 1 -hop neighbor3. Recover the affected nodes through 2 -hop neighbor4. Reconstruction of the whole tree

In our proposed route recovery method, if the sender node detects the linkfailure along the routing tree, the sender node try to recover the route usingthe other interfaces; if this is impossible then the sender node just broadcastthe error message to 1 -hop neighbors. A 1 -hop neighbor that can communi-cate with the affected receiver, response back with the reply message; a 1-hopneighbor that cannot cover the affected receiver, does not send any reply.Thereafter, if the sender of error message does not receive any reply after atimeout interval; it sends the error message to its 2 -hop neighboring nodesto recover the route from two level higher in the broadcast routing tree. Inmost of the scenarios, the route is recovered at this stage. If the route is stillnot recovered, then the sender node sends the request to the source node forreconstruction of the broadcast routing tree.

5 Network Model

We build and solved our research problem in configuration setting of multi-channel multi-radio CRNs. We take the following assumptions: (a) CRN nodeshave multiple radio interfaces, (b) CRN has multiple channels, (c) conditions ofthe network are dynamic, and (d) PU activity in unknown a priori. We assumethat interference is generated either by PU activity that is considered as a spe-cific source of the interference in CRNs or by a general interference source [42].Moreover, we are assuming one channel is dedicated to transmit the controlinformation. A node tunes the control channel on one of its interface. Wealso assume that channel assignment algorithm manages the channel switch-ing and notifies the transmission quality-aware broadcast routing algorithm.Generally, when an interferer originates the interference, the communicationin that proximity of the network suffers. In a common wireless network, thesignal is successfully received on a receiver when SINR value for that signalis sufficiently high. However, in CRNs, the PU activity obstructs the networkcommunication of secondary users. In CRN literature, the location of a PUand the PU arrival time are modeled by probabilistic distribution. The activityof PUs is uniformly distributed in network space and among various channels.However, the activity of PUs in terms of time is mathematically modeled byan ON/OFF Markov Renewal Process (MRP) [43, 44]. The ON/OFF MRPhas extensively used by various researchers in the CRN literature [43–48].The authors in [49] have estimated and validated the PU ON/OFF MRPfor the PU signal in WLAN IEEE 802.11b. Moreover, the ON/OFF MRPmodel is also used to approximate the spectrum usage pattern of public safetybands [50] [51].


6 Performance Evaluation

The performance of transmission-quality-aware broadcast routing frameworkis evaluated by simulating it in OMNeT++. We take a network area of 900× 900 m2 in simulator space. We have deployed 10 secondary users in theCRN for evaluating the performance of the algorithm. We analyze the per-formance of the transmission-quality-aware broadcast routing framework invarious configurations of CRN. The PU activity in time space is modeled asan ON/OFF MRP. Moreover, the ON and OFF periods are modeled by theexponential distribution as in [52]. The activities of PUs and SUs are uni-formly distributed among various channels. The rate parameters λX and λYof exponential distribution are taken as measured and presented in [52]. ThePROBE packets send time and PLR compute time have taken as 3 secondsand 30 seconds after simulating with different time intervals (1, 1.5, 2, 2.5, 3,3.5, 4, 4.5, and 5 seconds for PROBE packets send time) and (25, 26, 27, 28,29, 30, 31, 32, 33, 34, 35 seconds for PLR compute time). Moreover, the PUactivity detection is performed after every 1 second which is also selected aftersimulating different options as follows: 200 milliseconds, 400 milliseconds, 600milliseconds, 800 milliseconds, 1000 milliseconds, 1200 milliseconds, 1400 mil-liseconds, 1600 milliseconds, 1800 milliseconds, 2000 milliseconds. The valuesother than 1000 milliseconds do not perform optimally because of the tradeoffthat exists between detection accuracy and overhead. The lower values of thePU activity detection reduce the spectrum utilization for data exchange whilehigher values aggravates PU misdetection probability. We found the foremen-tioned values optimal for our protocol. The results obtained with other timeintervals are excluded due to limited space. In our system, we assume that eachreceiver already knows the rate at which transmitter send PROBE packets,which can be learned either by training or by manual configuration. The PLRon a receiver for each transmitter is computed for all its channels that makelinks between the node pair.

We consider that each secondary node has three radio interfaces. The sim-ulation runs for 1000 seconds of simulation time and is repeated for 30 runs.The results presented in the graphs are an average of the parameter output.The observed parameter values in the start for five seconds are discarded andare excluded when compiling results to eliminate the initialization bias. Thetransmission quality is computed by active exchange of PROBE messages dur-ing the initial phase, thereafter, the quality is passively measured on a tunedchannel.

To evaluate the performance of proposed algorithm, we focus on two pa-rameters: a) packet delivery ratio and b) throughput. In evaluation of our pro-posed framework, we have measured the performance by (a) changing numberof PUs in the wireless network, (b) changing number of available channels inthe system, and (c) changing PU activity level, that is modeled by the ONduration of the ON-OFF renewal process.


6.1 Packet Delivery Ratio

Packet delivery ratio (PDR) is the average number of packets received by allthe receivers. The graphs in Figures 7, 8, and 9 show the effect of number ofPUs on PDR for 3, 5 and 8 channels, respectively.

Fig. 7 Area = 900 × 900 m2, Number of SUs = 10, λOFF = 5, λON = 1.5, Number ofradios = 3, Number of Channels = 3

Packet delivery ratio for all three algorithms has no significant differencewhen there is no PU. Transmission quality aware (TQ-aware) broadcast rout-ing algorithm maintains higher PDR, compared to TQ-unaware broadcastrouting and MRDT. In addition, packet delivery ratio gradually decreases incase of TQ-aware broadcast routing algorithm, compare to drastic reductionfor TQ-unaware broadcast and MRDT. The study shows the number of PUnegatively affects the PDR of broadcast traffic in CRNs. However, the compar-ison of the graphs in Figures 7, 8, and 9 show that decrease in packet deliveryratio is also affected by number of channels in the network. Packet delivery ra-tio with 3 channels in the network is significantly lower than that of networkswith 5 and 8 channels. The reason behind the lower packet delivery ratio inthe network with 3 channels is higher occupancy probability by the PU. Insummary, our proposed algorithm outperforms the alternative algorithms weconsider even when the number of available primary channels are less.




6.2 Throughput

We define throughput as the average number of packets successfully receivedby destination nodes during the whole simulation period.


Fig. 10 Area = 900 × 900 m2, Number of SUs = 10, λON = 1.5, λOFF = 5, Number ofradios = 3, Number of Channels = 3


The results in Figures 10, 11, and 12 show the effect of number of PUs onthroughput for 3, 5, and 8 channels, respectively. Throughput for all three algo-


rithms has no significant difference when there is no PU. Transmission qualityaware (TQ-aware) broadcast routing algorithm maintains higher throughput,compared to TQ-unaware broadcast routing and MRDT. In addition, through-put decreases gradually in case of TQ-aware broadcast routing algorithm, com-pare to drastic reduction for TQ-unaware broadcast routing and MRDT.


The figures show that the number of PU negatively affects the through-put of broadcast traffic in CRNs. However, the comparison of the graphs inFigures 10, 11, and 12 show that decrease in throughput is also affected bynumber of channels in the network. Throughput in the network with 3 chan-nels is significantly lower than that of networks with 5 and 8 channels. Thereason behind the lower throughput in the network with 3 channels is higheroccupancy probability by the PU. In short, it is noticed that the TQ-awareoutperforms when the primary channels are limited.

6.3 Impact of varying available number of channels

Figure 13 shows the impact of number of available primary channels on packetdelivery ratio. The OFF duration λ is taken as 5 and the number of PUsis considered to be 25. The result shows that proposed algorithm performsbetter than its counterparts even when the number of channels is less. Theperformance of all the algorithms highly differs for equal number of channelsand interfaces. The availability of larger set of primary channels in network


Fig. 13 Area = 900 × 900 m2, Number of PUs = 25, Number of SUs = 10, λOFF = 5,λON = 1.5, Number of radios = 3

Fig. 14 Area = 900 × 900 m2, Number of PUs = 25, Number of SUs = 10, λON = 1.5,λOFF = 5, Number of radios = 3


lessens the probability of simultaneous transmissions by PUs and secondaryusers. Therefore, performance difference of the algorithms reduces with theincrease in number of available primary channels.

The throughput performance also ameliorates with the increasing numberof available primary channels in a CRN. In Figure 14, the impact of changingthe number of available primary channels is shown for throughput. The PUidle duration of the ON-OFF renewal process is taken as 5 and number ofPUs for this study are taken as 25. The simulation results present that thetransmission-quality-aware broadcast routing algorithm performs better thanother two schemes even when number of channels is less. The performancedifference for all the algorithms is greater when the number of channels is less.Hence, the performance of all broadcast routing solutions degrades with the in-crease in number of primary channels because of low probability of concurrenttransmissions by primary and secondary users.

6.4 PU Activity Ratio Impact

This subsection presents the analysis of the PU activity ratio impact on theperformance of secondary user application such as throughput and packetdelivery ratio.

Fig. 15 Area = 900 × 900 m2, Number of PUs = 15, Number of SUs = 10, λON = 1.5,Number of radios = 3, Number of channels = 5

The impact of PU activity ratio on PDR is depicted in Figure 15. Forevaluating the impact of PU activity ratio, the number of PUs is taken to


Fig. 16 Area = 900 × 900 m2, λON = 1.5, Number of PUs = 15, Number of SUs = 10,Number of radios = 3, Number of channels = 5

be 15, the system has 5 primary channels, and each secondary user device isequipped with three radios.

The results show that transmission-quality-aware broadcast routing frame-work perform better than its counterpart algorithms even when PU activityis more. The result shows that packet delivery ratio and throughput decreasewith the increase in PU activity. The more OFF duration of PUs in networkreduces the probability of number of simultaneous transmissions by PUs andsecondary users that results in higher packet delivery ratio and throughput.The packet delivery ratio and throughput for TQ-aware broadcast routingalgorithm is higher even for higher PU activity ratio.

6.5 Effect of Transmission Quality Enhancement Phases on Packet DeliveryRatio

We have also studied the impact of different phases on improvement in per-formance of broadcast routing.

The broadcast performance in terms of packet delivery ratio is measured byvarying number of PUs in the network for different combinations of transmission-quality enhancement phases: without transmission quality enhancement (With-outQualityEnhancement), transmission quality enhancement intra-node only(TQIntraNodeEnhancement), transmission quality enhancement inter-node only(TQInterNodeEnhancement), transmission quality enhancement intra-node andinter-node both (TQIntraandInterNodeEnhancement).


Fig. 17 Area = 900 × 900 m2, λOFF = 5, λON = 1.5, Number of SUs = 10, Number ofradios = 3, Number of channels = 5

The result presented in Figure 17 shows a significant improvement in packetdelivery ratio when the quality is enhanced internally on a node and exter-nally between the nodes in broadcast routing. The difference in performanceimprovement is increased when the number of users are higher. The maximumincrease in performance is observed to be 17% when the PDR for includingboth quality enhancement phases is 0.65 with 25 PUs in the network andthe PDR without transmission quality enhancement is observed as 0.49. Theimprovement in PDR is ensured by migrating the bottleneck receiver nodeseither to other interfaces of the same node, or to other node’s interfaces.

7 Conclusions

We have proposed an adaptive broadcast routing framework that employs anovel formulation of ‘transmission quality’ in the form of transmission-quality-aware routing metric. The transmission quality-aware broadcast routing metricincorporates the packet loss ratio and primary user activity ratio on the linkin the route selection process. This paper also presents a transmission quality-aware broadcast routing framework that employs the transmission quality-aware broadcast routing metric. We evaluated the performance of proposedrouting framework by simulating the framework in OMNeT++. The sim-ulation results show that the transmission quality-aware broadcast routingframework performs better than existing broadcasting work with more inter-


ference/ or primary user arrivals and attained better packet delivery ratio andthroughput. The throughput and packet delivery ratio for proposed routingframework are 40% better than transmission quality un-aware routing frame-work and MRDT. It is observed that packet delivery ratio and throughputdecrease gradually in case of TQ-aware broadcast routing algorithm, compareto drastic reduction faced by TQ-unaware broadcast routing algorithm andMRDT. Furthermore, the increase in number of primary users and their ac-tivity ratio also negatively affects the packet delivery ratio and throughput ofbroadcast traffic in cognitive radio networks. Lastly, it is noticed that the in-crease in number of channels also positively affect the throughput and packetdelivery ratio as it reduces the probability of concurrent transmission by pri-mary users and secondary users in cognitive radio networks.

Acknowledgments

This work is supported in part by the H.E.C. Pakistan and Malaysian Ministryof Higher Education, as the University of Malaya High Impact Research Grant(UM.C/625/1/HIR/MOE/FCSIT/03).

References

1. S. Abolfazli, Z. Sanaei, E. Ahmed, A. Gani, and R. Buyya, “Cloud-based augmentationfor mobile devices: motivation, taxonomies, and open challenges,” IEEE CommunicationsSurveys & Tutorials, vol. 16, no. 1, pp. 1–32, 2014.

2. M. Sookhak, H. Talebian, E. Ahmed, A. Gani, and M. K. Khan, “A review on remotedata auditing in single cloud server: Taxonomy and open issues,” Journal of Network andComputer Applications, vol. 43, pp. 121–141, 2014.

3. E. Ahmed, A. Akhunzada, M. Whaiduzzaman, A. Gani, S. H. Ab Hamid, and R. Buyya,“Network-centric performance analysis of runtime application migration in mobile cloudcomputing,” Simulation Modelling Practice and Theory, 2014.

4. M. Whaiduzzaman, M. Sookhak, A. Gani, and R. Buyya, “A survey on vehicular cloudcomputing,” Journal of Network and Computer Applications, vol. 40, pp. 325–344, 2014.

5. X. Yu and F. Sun, “Intelligent urban emergency early warning system based on dynamicrough set and cloud computing,” in Proc. of 4th IEEE International Conference on Soft-ware Engineering and Service Science (ICSESS’13), Haidian Dist. Beijing, China. IEEE,2013, pp. 701–704.

6. C. Busch, R. Kannan, and A. V. Vasilakos, “Approximating congestion+ dilation innetworks via quality of routing games,” IEEE Transactions on Computers, vol. 61, no. 9,pp. 1270–1283, 2012.

7. A. Cianfrani, V. Eramo, M. Listanti, M. Polverini, and A. V. Vasilakos, “An ospf-integrated routing strategy for qos-aware energy saving in ip backbone networks,” IEEETransactions on Network and Service Management, vol. 9, no. 3, pp. 254–267, 2012.

8. P. Li, S. Guo, S. Yu, and A. V. Vasilakos, “Codepipe: An opportunistic feeding androuting protocol for reliable multicast with pipelined network coding,” in Proc. of IEEEINFOCOM. IEEE, 2012, pp. 100–108.

9. P. Basarkod and S. Manvi, “Node movement stability and congestion aware anycastrouting in mobile ad hoc networks,” in Proc. of IEEE International Advance ComputingConference (IACC’14), Queensway, Hong Kong, Feb 2014, pp. 124–131.

10. Z. Zhao, T. Braun, D. Rosario, and E. Cerqueira, “CAOR: Context-aware adaptiveopportunistic routing in mobile ad-hoc networks,” in Proc. of 7th IFIP Wireless andMobile Networking Conference (WMNC’14), Vilamoura, Algarve, Portugal, May 2014,pp. 1–8.


11. Y.-S. Yen, H.-C. Chao, R.-S. Chang, and A. Vasilakos, “Flooding-limited and multi-constrained QoSmulticast routing based on the genetic algorithm for manets,” Mathemat-ical and Computer Modelling, vol. 53, no. 11, pp. 2238–2250, 2011.

12. X. Wang, A. V. Vasilakos, M. Chen, Y. Liu, and T. T. Kwon, “A survey of green mobilenetworks: Opportunities and challenges,” Mobile Networks and Applications, vol. 17, no. 1,pp. 4–20, 2012.

13. S. Kalantari, Z. S. Daliri, S. Shamshirb, L. S. Ng et al., “Routing in wireless sensornetwork based on soft computing technique,” Scientific Research and Essays, vol. 6, no. 21,pp. 432–4441, 2011.

14. M. Li, Z. Li, and A. V. Vasilakos, “A survey on topology control in wireless sensornetworks: Taxonomy, comparative study, and open issues,” Proceedings of the IEEE, vol.101, no. 12, 2013.

15. Y. Yao, Q. Cao, and A. V. Vasilakos, “Edal: an energy-efficient, delay-aware, andlifetime-balancing data collection protocol for wireless sensor networks,” in IEEE 10thInternational Conference on Mobile Ad-Hoc and Sensor Systems (MASS’13). IEEE,2013, pp. 182–190.

16. L. Xiang, J. Luo, and A. Vasilakos, “Compressed data aggregation for energy efficientwireless sensor networks,” in Proc. of 8th Annual IEEE Communications Society Confer-ence on Sensor, Mesh and Ad Hoc Communications and Networks (SECON’11). IEEE,2011, pp. 46–54.

17. Y. Zeng, K. Xiang, D. Li, and A. V. Vasilakos, “Directional routing and scheduling forgreen vehicular delay tolerant networks,” Wireless networks, vol. 19, no. 2, pp. 161–173,2013.

18. T. Spyropoulos, R. N. Rais, T. Turletti, K. Obraczka, and A. Vasilakos, “Routing fordisruption tolerant networks: taxonomy and design,” Wireless networks, vol. 16, no. 8, pp.2349–2370, 2010.

19. A. V. Vasilakos, Y. Zhang, and T. Spyropoulos, Delay tolerant networks: Protocols andapplications. CRC Press, 2012.

20. J. Qadir, “Artificial intelligence based cognitive routing for cognitive radio networks,”arXiv preprint arXiv:1309.0085, 2013.

21. A. Abbagnale and F. Cuomo, “Connectivity-driven routing for cognitive radio ad-hocnetworks,” in Proc. of 7th Annual IEEE Communications Society Conference on SensorMesh and Ad Hoc Communications and Networks (SECON’10), Boston, Massachusetts,USA. IEEE, 2010, pp. 1–9.

22. M. Youssef, M. Ibrahim, M. Abdelatif, L. Chen, and A. Vasilakos, “Routing metrics ofcognitive radio networks: A survey,” IEEE Communications Surveys Tutorials, vol. 16,no. 1, pp. 92–109, First 2014.

23. E. Ahmed, M. Shiraz, and A. Gani, “Spectrum-aware distributed channel assignment forcognitive radio wireless mesh networks,” Malaysian Journal of Computer Science, vol. 26,no. 3, pp. 232–250, 2013.

24. Y. Saleem, A. Bashir, E. Ahmed, J. Qadir, and A. Baig, “Spectrum-aware dynamicchannel assignment in cognitive radio networks,” in Proc. of International Conference onEmerging Technologies (ICET’12), Islamabad Pakistan. IEEE, 2012.

25. E. Ahmed, L. J. Yao, M. Shiraz, A. Gani, and S. Ali, “Fuzzy-based spectrum handoffand channel selection for cognitive radio networks,” in Proc. of International Conferenceon Computer, Control, Informatics and Its Applications (IC3INA’13). IEEE, 2013, pp.23–28.

26. A. K. Mir, A. Akram, E. Ahmed, J. Qadir, and A. Baig, “Unified channel assignmentfor unicast and broadcast traffic in cognitive radio networks.” in LCN Workshops, 2012,pp. 799–806.

27. M. Hassan, E. Ahmed, J. Qadir, and A. Baig, “Quantifying the multiple cognitive radiointerfaces advantage,” in Proc. of 27th International Conference on Advanced InformationNetworking and Applications Workshops (WAINA), 2013, March 2013, pp. 511–516.

28. H. Cheng, N. Xiong, A. V. Vasilakos, L. Tianruo Yang, G. Chen, and X. Zhuang, “Nodesorganization for channel assignment with topology preservation in multi-radio wirelessmesh networks,” Ad Hoc Networks, vol. 10, no. 5, pp. 760–773, 2012.

29. E. Ahmed, A. Gani, S. Abolfazli, L. Yao, and S. Khan, “Channel assignment algorithmsin cognitive radio networks: Taxonomy, open issues, and challenges,” IEEE Communica-tions Surveys & Tutorials, in press, 2014.


30. C. T. Chou, A. Misra, and J. Qadir, “Low-latency broadcast in multirate wireless meshnetworks,” Selected Areas in Communications, IEEE Journal on, vol. 24, no. 11, pp.2081–2091, 2006.

31. J. E. Wieselthier, G. D. Nguyen, and A. Ephremides, “Energy-efficient broadcast andmulticast trees in wireless networks,” Mobile Networks and Applications, vol. 7, no. 6, pp.481–492, 2002.

32. D. S. De Couto, D. Aguayo, J. Bicket, and R. Morris, “A high-throughput path metricfor multi-hop wireless routing,” Wireless Networks, vol. 11, no. 4, pp. 419–434, 2005.

33. J. Qadir, A. Misra, and C. T. Chou, “Minimum latency broadcasting in multi-radiomulti-channel multi-rate wireless meshes,” in Sensor and Ad Hoc Communications andNetworks, 2006. SECON’06. 2006 3rd Annual IEEE Communications Society on, vol. 1.IEEE, 2006, pp. 80–89.

34. J. Qadir, C. T. Chou, A. Misra, and J. G. Lim, “Minimum latency broadcasting in multi-radio, multichannel, multirate wireless meshes,” IEEE Transactions on Mobile Computing,vol. 8, no. 11, pp. 1510–1523, 2009.

35. W. Lou and J. Wu, “On reducing broadcast redundancy in ad hoc wireless networks,” inSystem Sciences, 2003. Proceedings of the 36th Annual Hawaii International Conferenceon. IEEE, 2003, pp. 10–pp.

36. J. Qadir, C. T. Chou, A. Misra, and J. G. Lim, “Localized minimum-latency broadcast-ing in multi-radio multi-rate wireless mesh networks,” in Proc. of International Symposiumon a World of Wireless, Mobile and Multimedia Networks, (WoWMoM’08) , NewportBeach, CA, USA. IEEE, 2008, pp. 1–12.

37. Y. R. Kondareddy and P. Agrawal, “Selective broadcasting in multi-hop cognitive radionetworks,” in IEEE Sarnoff Symposium. IEEE, 2008, pp. 1–5.

38. M. H. Rehmani, A. C. Viana, H. Khalife, S. Fdida et al., “Surf: A distributed channelselection strategy for data dissemination in multi-hop cognitive radio networks,” ComputerCommunications, vol. 36, no. 10, 2013.

39. V. Borges, M. Curado, and E. Monteiro, “Cross-layer routing metrics for mesh networks:Current status and research directions,” Computer Communications, vol. 34, no. 6, pp.681–703, 2011.

40. X. Zhao, J. Guo, C. T. Chou, A. Misra, and S. Jha, “A high-throughput routing metricfor reliable multicast in multi-rate wireless mesh networks,” in Proc. of IEEE InternationalConference on Computer Communications (INFOCOM’11), Shanghai, China. IEEE,2011, pp. 2042–2050.

41. J. Qadir, A. Baig, A. Ali, and Q. Shafi, “Multicasting in cognitive radio networks:Algorithms, techniques and protocols,” Journal of Network and Computer Applications,vol. 45, pp. 44–61, 2014.

42. P. Cardieri, “Modeling interference in wireless ad hoc networks,” IEEE CommunicationsSurveys & Tutorials, vol. 12, no. 4, pp. 551–572, 2010.

43. G. Yuan, R. C. Grammenos, Y. Yang, and W. Wang, “Performance analysis of selectiveopportunistic spectrum access with traffic prediction,” IEEE Transactions on VehicularTechnology, vol. 59, no. 4, pp. 1949–1959, 2010.

44. A. W. Min and K. G. Shin, “Exploiting multi-channel diversity in spectrum-agile net-works,” in Proc. of The 27th IEEE Conference on Computer Communications. (INFO-COM’08), Phoenix, AZ, USA. IEEE, 2008, pp. 1921–1929.

45. W.-Y. Lee and I. F. Akyildiz, “Optimal spectrum sensing framework for cognitive radionetworks,” IEEE Transactions on Wireless Communications, vol. 7, no. 10, pp. 3845–3857,2008.

46. H. Kim and K. G. Shin, “Fast discovery of spectrum opportunities in cognitive radionetworks,” in Proc. of 3rd IEEE Symposium on New Frontiers in Dynamic SpectrumAccess Networks, (DySPAN’08) , Chicago, Illinois USA. IEEE, 2008, pp. 1–12.

47. O. Mehanna, A. Sultan, and H. E. Gamal, “Cognitive mac protocols for general primarynetwork models,” arXiv preprint arXiv:0907.4031, 2009.

48. A. S. Zahmati, X. Fernando, and A. Grami, “Steady-state markov chain analysis forheterogeneous cognitive radio networks,” in IEEE Sarnoff Symposium, 2010. IEEE,2010, pp. 1–5.

49. S. Geirhofer, L. Tong, and B. M. Sadler, “Dynamic spectrum access in WLAN channels:empirical model and its stochastic analysis,” in Proc. of the first international workshopon Technology and policy for accessing spectrum. ACM, 2006, p. 14.


50. L. Yang, L. Cao, and H. Zheng, “Proactive channel access in dynamic spectrum net-works,” Physical Communication, vol. 1, no. 2, pp. 103–111, 2008.

51. B. Vujicic, “Modeling and characterization of traffic in a public safety wireless network,”Ph.D. dissertation, Simon Fraser University, 2006.

52. H. Kim and K. G. Shin, “Efficient discovery of spectrum opportunities with mac-layersensing in cognitive radio networks,” IEEE Transactions on Mobile Computing, vol. 7,no. 5, pp. 533–545, 2008.

High Throughput Transmission Quality Aware Broadcast Routing in Cognitive Radio Networks

Documents