Link prediction-based topology control and adaptive routing in Cognitive Radio Mobile Ad-Hoc Networks

Abstract

Cognitive Radio (CR) is a promising technology which deals with using vacant spectrum of licensed frequency band opportunistically. In CR Network, route construction must not affect the transmission of Primary User activity. In CR technology challenge of maintaining optimal routes arises due to PU activity and mobility of spectrum resources i.e. CR users. Maintaining optimal routes in Ad-hoc CR network increases the overall network throughput and decreases end-to-end delay. This end-to-end network performance can be increased by providing cognition capability to routing in Cognitive Radio Mobile Ad-Hoc Network. We propose routing scheme with link-availability prediction and topology control. Link-availability prediction considers primary user activities and user mobility. Topology control constructs dynamic and reliable topology based on link prediction. This routing scheme reduces energy consumption, rerouting and thus enhances overall network performance and ensures least delay.

Full text

Link Prediction-Based Topology Control and
Adaptive Routing in Cognitive Radio Mobile Ad-Hoc
Networks
Kanchan Hadawale
Computer Department
MIT Academy of Engineering
Alandi(D), Pune, India
Kanchan.hadawale@yahoo.in
Sunita Barve
Computer Department
MIT Academy of Engineering
Alandi(D), Pune, India
Barve_sunita@reddifmail.com
Abstract—Cognitive Radio (CR) is a promising technology
which deals with using vacant spectrum of licensed frequency
band opportunistically. In CR Network, route construction must
not affect the transmission of Primary User activity. In CR
technology challenge of maintaining optimal routes arises due to
PU activity and mobility of spectrum resources i.e. CR users.
Maintaining optimal routes in Ad-hoc CR network increases the
overall network throughput and decreases end-to-end delay. This
end-to-end network performance can be increased by providing
cognition capability to routing in Cognitive Radio Mobile Ad-
Hoc Network. We propose routing scheme with link-availability
prediction and topology control. Link-availability prediction
considers primary user activities and user mobility. Topology
control constructs dynamic and reliable topology based on link
prediction. This routing scheme reduces energy consumption,
rerouting and thus enhances overall network performance and
ensures least delay.
Keywords—Cognitive Radio Network (CRN); Cognition
Capability; Link Prediction; Topology Control
I. INTRODUCTION
Cognitive Radio (CR) technology is one of promising
technology that allows unlicensed users to access licensed
spectrum bands opportunistically in a dynamic and non-
interfering manner. Federal Communication Commission
(FCC) [1] highlights that many licensed spectrum bands are
used only in specific geographical areas and over limited
periods of time where average utilization of spectrum band
varies between 15% and 85%. Cognitive radio technology is
thus introduced to solve the problem of spectrum usage
inefficiency.
In CRN most research is done on MAC layer issues such as
opportunistic spectrum access and spectrum utilization.
Whereas CR technology has more impact on upper layer
performance issues in wireless network, specifically in Mobile
Ad-Hoc Networks (MANETs). In routing issue, data should be
routed via stable and reliable path to avoid rerouting and thus
congestion in network. This degrades the performance of
network such as throughput and delay. As compared to
classical routing, routing in CRN is more unstable as it is
affected by not only mobility of Cognitive User’s (CU) but also
by Primary User (PU) activities. Thus routing in CR-MANETs
should have the following characteristics to ensure stable and
reliable paths.
a) PU activity awareness: In CRN path selection should be
in such a way that it should not interfere to PU activity i.e.
PU interference should be below the required threshold.
b) Link-availability: To avoid PU interference, cognitive
routing should also be reactive. It should be aware of link-
available periods.
c) Adaptive behavior: Cognitive routing should be adaptive
to selected path based on some prediction to avoid
rerouting frequency and to increase end-to-end throughput
and to decrease end-to-end delay.
There exists large no. of routing protocols for MANETs but
because of distinct characteristics of CR-MANETs, that
protocols cannot be apply directly to CR-MANETs.
So here we proposed technique that provides cognition
capability to routing through middleware mechanism like
topology control. Topology control is used in almost all
wireless networks to reduce energy consumption and
interference. With topology control network connectivity is
done on link information provided by MAC and physical
layers. Thus topology control works as middleware that
connects routing and lower layers. In CRMANETs, topology
control takes care of PU activities and spectrum-availability.
In proposed scheme to provide cognition capability to routing,
prediction based topology control is considered. This
technique captures topology dynamically based on the link
information provided by lower layers to provide opportunistic
link management and routing in CR-MANETs. Link-
prediction model is proposed to deal with cognitive user
mobility and PU activities. This model predicts the link-
availability duration as well as probability that duration
remains till the end of this period. Based on this predicted
links, reliable topology is constructed to reduce rerouting. For
link prediction, local neighbor information is collected and
network connectivity is preserved in distributed manner.
2014 IEEE Global Conference on Wireless Computing and Networking (GCWCN)
978-1-4799-6298-3/14/$31.00 © 2014 IEEE
11

II. RELATED WORK
A. Link Stability Routing Solutions
Link availability in cognitive network is different from
traditional wireless network. Link availability in cognitive
network varies with time and space. Thus stable links is one of
the routing solutions which can be achieved through following
routing schemes.
1) Enhanced Path Recovery Functionalities
In [2], authors proposed SPEctrum-Aware Routing
(SPEAR) protocol to get link flexibility with spectrum
heterogeneity. Spectrum availability depends on location and
presence of primary user activities. SPEAR considers
following concepts for link stability,
•To deal with spectrum heterogeneity, spectrum
discovery is integrated with route discovery.
•By minimizing inter-flow interference, channel
assignments are coordinated on per-flow basis.
•To achieve spectrum diversity and to reduce intra-
flow interference, local spectrum heterogeneity is
considered.
SPEAR set-up route by broadcasting and AODV-style route
discovery is used to get each node’s channel quality and
availability information. Each Route Request (RREQ) contains
node IDs, nodes spectrum availability and link quality. All
these parameters are combined at destination to select optimal
route. SPEAR discovers multiple paths with redundant paths
that are sent to destination for best path selection. Selected
route is reserved at destination by using RREP messages. In
[3], collaboration between route selection and spectrum
decision is considered. In this paper, authors proposed
Spectrum Tree based On Demand routing Protocol (STOD-
RA) which includes,
•Route metric based on PUs activities and SUs QoS
requirements.
•Spectrum-tree structure of each available channel
•Spectrum Tree based On Demand Routing Algorithm
Routing metric combines both link stability and spectrum
availability. It predicts spectrum availability time from history
of PU activities.
2) Targetting Route Stability
In [4], link stability is associated to path connectivity through
mathematical model. Degree of connectivity is considered
while measuring paths, also PU behavior is also considered.
Authors introduce novel metric to assign weight to routes
which also captures path stability and availability over time. In
this paper routing scheme is named as “Gymkhana”, which
forwards information along paths to avoid unstable and low
connectivity network zones. Gymkhana uses distributed
protocol to collect information of candidate paths from source
to destination. Gymkhana contributes following work in
cognitive routing,
•Provision of re-elaboration of algebraic connectivity
in cognitive context.
•Formulation of path connectivity and path length
used in cognitive routing protocol.
In [5], route stability is defined In terms of route maintenance
cost and protocol is designed. Route maintenance cost
represents efforts required for maintaining end-to-end
connectivity in cognitive routing. It includes operations like
link switching and channel switching on presence of primary
user activity. In link switching link routes must be replaced by
the links that are not affected by PU. In channel switching
same link can be maintained but spectrum portion must be
changed. Authors propose MILP formulation to minimize
route maintenance cost with link capacity and flow balance
constraints. Centralized algorithm is designed to compute
minimum maintenance cost routes in cognitive routing.
3) Routing with Mobile SUs
In [6], SEARCH routing protocol is designed on geo-
graphic forwarding principle. In cognitive radio network, route
is constructed at network layer must not affect primary user’s
transmission and thus must be aware of spectrum availability.
The frequency changing PU activity and mobility of CR user
make the problem of maintaining optimal routes in Ad-Hoc
cognitive radio network challenging. SEARCH mainly works
on following two concepts,
a) PU activity awareness: In CR network, route must be
constructed to avoid region affected by active PU.
When PU activity affect region, SEARCH provides
hybrid solution, it first uses greedy geographic routing
on each channel to reach destination by identifying
and circumventing PU activity region. The path
information from different channels is combined at
destination in series of optimization steps to decide on
optimal end-to end route in a computationally
efficient way.
b) CR user mobility: Cognitive user mobility results into
frequent route disconnections. Thus for each node,
through periodic beacons, updates its one-hop
neighbor about it current location. SEARCH ensures
performance as well as less interference in cognitive
radio network.
B. Solutions with Link-Availability Prediction Model
In [7], to reduce rerouting, path availability and reliability
metric is concerned in routing of MANETs. For that
prediction based link-availability estimation and path
reliability estimation is used as routing metric. The basic idea
of this estimation is nodes first predict that a continuous time
period (Tp) that the link-availability remains from t0 with
assumption that velocity both nodes of that link will be
constant during Tp. Then the probability estimation L(Tp),
that this link remains until t0 + Tp, with possible changes in
velocities of nodes during t0 to t0 + Tp. Thus with these link-
availability estimation they get two approaches, “unaffected
Tp” with constant velocities and “affected Tp” with changes in
velocities. This estimation improves the tendency of link
availability to develop path selection metrics to improve
network performance.
In [8], authors proposed enhancement to the prediction based
link availability estimation. Since, in original estimation L (Tp)
all possible changes in velocities during Tp do not take into
12

account. It only considers first change in velocity during time
period Tp and link availability for these changes is estimated
by . This is given as,
L (Tp) ≈ ()2
11
+
22
Tp
eTp
Tp Tp
λρλ
λλ
⎛⎞
⎡⎤ ⎡⎤
⎜⎟
⎢⎥ ⎢⎥
⎜⎟
⎣⎦ ⎣⎦
⎝
−
+−−
⎠
εε
Where, p = probability of two nodes of a link to get close to
each other. For link prediction Tp, node can measure the link
duration that remains till the end of Tp and link-availability
measured as Lm (Tp) = (Tr) / (Tp).
III. SYSTEM MODEL
Cognitive Radio Network (CRN) routing is different from
classical wireless routing in terms of dynamic spectrum
availability and PU activity interference. Spectrum availability
is affected by following two factors,
a) PU activities: As CUs are considered as secondary
user due to low priority in accessing the spectrum
band allocated to PU. CU should be aware of PU
activities while sensing the spectrum. Spectrum
sensing can be done either non-cooperatively i.e.
individually each CU conducts radio detection and
takes decision by itself or cooperatively where
spectrum sensing is done by group of CUs in
collaboration.
b) CU mobility: Due to node mobility frequent
disconnection occurs in established route. Route
disconnection is detected when next-hope node in
route does not reply to message and retry limit is
exceeded. In CR-MANETs, due to node mobility in
PU activity region spectrum also becomes
unavailable.
To solve the spectrum availability problem, above two factors
should be consider for designing efficient topology control
and routing schemes. In proposed system, focus is on
spectrum availability affected by CU mobility. That means
cognitive routing should select the links with long path
survival time to improve path stability. This cognition
capability is achieved through middleware technology such as
topology control. Topology control works with cross-layer
module that connects routing layer and CR module. In
proposed scheme topology control is integrated with link-
availability prediction. The proposed system is divided into 3
stages,
Fig. 1 Proposed system architecture
A reasonable link-availability prediction model is used for
proposed system. The basic principle is to provide predicted
time period Tp for link-availability between two nodes. L(Tp)
parameter is used to estimate probability that this link may
remain till the end of Tp by considering changes in velocities.
It also estimates L(Tp) for a node that does not move at a
constant velocity. In CRMANETs along with CU mobility PU
activity region is considered for link prediction. To avoid PU
interference distance between CU and PU is calculated. For
link-availability before node moves into PU activity region,
another pair of parameters [Tp,L(Tp)] is calculated. is
predicted time till CU is outside the PU activity region and
L( ) is its corresponding probability. Final link prediction
we get by combining [Tp, L(Tp)] and [ , L( )].
1) [, L ( )] Estimation
As we consider that the velocities of nodes stay constant
during random long period of times, so it is easy to get
(δ2d2)/(δ2T2)=0.
Where, d = distance between two nodes T = time interval
Then d can be calculated as,
d2 = αT2 + βT + γ (1)
Where, α, β and γ = constant
α, β and γ can be calculated by three points of measurements
(t0, d0), (t1 , d1) and (t2 , d2). Sample time is ti = t0 + Ti and di is
measured distance between two nodes. Without loss of
generality, Let t0 = 0 then α, β and γ can be sorted as,
α = [( t2 – t1) – (t2 – t1)] / [t1t2 (t1-t2)]
β = [( – ) – ( – )] / [t1t2 (t1-t2)]
γ =
When two nodes are in each other’s transmission range and
their velocities are constant and different from each other than
there is also possibility that they will travel out of this range
then Tp can be considered for equation 1. Some nodes always
remain outside the PU activity region then there is no solution
for equation 1. Once CU is detected in PU activity region
by CR module then we don’t need to predict link availability
thus, is set to 0. This system consider situation where
nodes are out of PU activity region.
Let, ρ = interference radius of PU then,
∆ = β2 + 4αρ2 - 4αγ
To get one solution for equation 1 of we have ∆ >= 0.
Consider >= 0, the available time period Tp if counted
from t2 is,
= 22
44
2
2t
βαραγ
α
⎧
⎪
⎪+−
−
⎨
⎪
⎪∞
⎩
if ∆ >= β2 (2)
Preciseness depends on ρ, we can be derived by propagation
model and measured received signal strength by time-of
arrivals based distance measurement for indoor environment
and GPS system for outdoor environment. Similarly
probability L ( ) to can be derived as,
L ( )
ˆˆ
(1 e )
Te T
pp
e
λτ
λλ
ζ
−
−−
≈+− (3)
Where, τ and ζ are obtained by measurements. Then the
pair [ , L ( )] can predict link duration corresponding to
PU interference.
2) Link-Availability Prediction
Link is considered to be available if two nodes associated
with link are within transmission range of each other and both
are out of interference region of any PU. We consider Tp * L
(Tp) routing metric to assist routing protocols in reliable path
13

selection. But if Tp * L (Tp) > * L ( ) then it suppress
transmission. Thus, link available duration Ta should be set to,
{
}
ˆˆ
min *L(T ),T * (T )
1,2, {PUs}
jj
TTL
pppi pi
aij
=
=∈ (4)
Where, i = associated with 2 ends of link
{PUs} = all PUs in network
i, j = indicates boundary of link i.e. if any end moves in PU
activity region then link becomes unavailable.
Thus, Ta is cognitive feature enabled which considers node
mobility and PU activity region.
A. Distributed Topology Construction
Based on link prediction, topology control and routing
scheme is proposed. In wireless network, topology control is
introduced to save energy consumption. In CR-MANETs due
to CU mobility and PU activity region causes frequent
rerouting and thus results into low end-to-end network
performance. Thus to solve this problem more reliable
topology is constructed in CR-MANETs.
New link reliability metric is defined for topology control.
In CR-MANETs rerouting penalty denoted by δ is the period
that occurred by rerouting is one of the metric. Rerouting
penalty reduces link availability to (Ta - δ). In CR-MANETs it
is requires fewer rerouting if selected path consists links of
longer Ta and higher quality. Link weight can be estimated as,
w = r * (Ta - δ)
Where, r = link data rate, δ = rerouting penalty
Here δ is converted into capacity loss r* δ during available
period. Traffic carrying ability of link is represented by link
weight. Path weight can be defined as,
W = min wi i  L
Where, L = all links along the path
In proposed scheme, focus is on to transmit more data traffic
before link failure. Topology construction is three step
process: Neighbor collection, path search and neighbor
selection. For CR-MANETs to preserve end-to-end reliable
paths, distributed cognitive algorithm uses enhancement of
Localized Dijkstra Topology Control (LDTC) algorithm. With
this distributed algorithm, each node performs following
functions as an initial node.
a) Neighbor Collection:
Exchange local information within all its neighbors
and calculate edge weight as, w = r * (Ta - δ)
b) Path Search:
For initial node set path weight to infinity and to zero
for neighbors. Mark initial node as current one and
all neighbors as unvisited.
Calculate path weight by using following equation,
from initial node to unvisited nodes. According to
change in path weights update the records.
W = min wi i  L
Then mark the current node as visited, and set the
unvisited node with largest path weight as the current
node. Until all nodes get visited repeat from step b).
c) Neighbor Selection:
Once all reliable paths from initial node are found,
select first hop neighbors from initial node along
these paths.
B. Routing on Resulting Topology
With link prediction, routing in CR-MANET considers
both CU mobility and PU activity region. It makes routing
adaptive to mobile environments by considering reliable paths.
For example, with DSR and AODV routing protocols, node
sends RREQ packets to find path from source to destination.
When RREQ reaches an intermediate node, it may drop if
transmitter does not exist in neighborhood relationship.
Otherwise this RREQ is disposed by intermediate node and
rebroadcast. Thus as a result, links in PU activity region or
poor quality are avoided. With predicted link durations,
performance of topology control can be evaluated.
IV. SIMULATION RESULTS
The proposed scheme is evaluated using computer network
simulator tool. In this random wireless environment nodes are
moving randomly in a 2-D space. IEEE 802.11 is used for
MAC layer. The performance evaluation of PCTC is plotted in
Fig.2. Simulation results are in area 600×500m2 with 45
mobile CUs. The original topology shows that the link
between 2 nodes exists whenever they are in each other’s
transmission range. In original topology PU awareness is not
considered. Whereas, in Fig.2 (b) and Fig.2 (c) shows that
with cognitive link duration Ta, links around PU are avoided
for communication. Thus the resulting topology makes
reliable paths available for communication.
The two performance metrics end-to-end delay and
throughput are also discussed on routing of predicted topology
and routing of prediction based topology control resulting
topology. We run simulation 50 times. Fig. 3 (a) shows
proposed topology has small average node degree and small
maximum node degree as compared to without topology
control mechanism. The small node degree simplifies route-
discovery process by reducing no. of RREQs and at same time
it also reduces contention in shared wireless medium. Fig. 3
(b) shows topology control algorithm also results into longer
link-durations. Thus the proposed topology provides cognition
capability to avoid interference to PUs activities, which in turn
provides adaptive routing over proposed topology in CR-
MANETs.
(a
)
14

(b)
(c)
Fig. 2 Topology comparison (a) Original Topology (b)
Topology with cognitive link-prediction (c) Prediction
Control Adaptive Topology Control resulting topology.
(a)
(b)
Fig. 3 Properties of resulting prediction based topology (a)
Node degree (b) Link duration
Conclusion
Proposed system provides cognition capability to routing
protocols in CR-MANETs which takes care of PU activities as
well as CU mobility. With last longer links, reliable topology is
constructed which makes protocol adaptive to mobility
environment. With resulting topology, rerouting is reduced
which increases overall end-to-end performance.
References
[1] I.F. Akyildiz, W. Y. Lee, M. Vuran, and S. Mohanty, “Next
generation/dynamic spectrum access/ cognitive radio wireless
networks:A survey,” Comput. Netw., vol. 50, no. 13, pp. 2127-2159,
Sep.2006
[2] A. Sampath, L. Yang, L. Cao, H. Zheng, B.Y. Zhao, “High throughput
spectrum-aware routing for cognitive radio based ad-hoc networks”, in:
3th International Conference on Cognitive Radio Oriented Wireless
Networks and Communications, Crwown-Com 2008
[3] G.-M. Zhu, I. Akyildiz, G.-S. Kuo, “STOD-RP: a spectrum-tree based
ondemand routing protocol for multi-hop cognitive radio networks”, in:
IEEE Global Telecommunications Conference, GLOBECOM 2008.
[4] A. Abbagnale, F. Cuomo, “Gymkhana: a connectivity- based routing
scheme for cognitive radio ad hoc networks”, in: IEEE Conference on
Computer Communications, INFOCOM 2010, pp. 1U5, doi:10.1109/
INFCOMW.2010.5466618.
[5] I. Filippini, E. Ekici, M. Cesana, “Minimum maintenance cost routing in
cognitive radio networks, in: Mobile Adhoc and Sensor Systems”,2009,
IEEE 6th International Conference on, 2009, pp. 284U293
[6] K. Chowdhury, M. Felice, “Search: a routing protocol for mobile
cognitive radio ad-hoc networks”, Computer Communications 32(18)
(2009) 1983-1997
[7] K. Yang and Y. Tsai, “Link stability prediction for mobile ad hoc
networks in shadowed environments, T in Proc. IEEE GLOBECOM,
San Francisco, CA, Nov. 2006, pp. 1U5
[8] S. Jiang, “An enhanced prediction-based link availability estimation for
MANETs,T IEEE Trans. Commun., vol. 52, no. 2, pp. 183U186, Feb.
2004.
[9] Quansheng Guan,Shengming Jiang,“Prdiction Based Topology Control
and Routing in Cognitive Radio Mobile Ad hoc Networks”, November
2010
[10] M. Cesana et al.,“Routing in cognitive radio networks: Challenges and
solutions”, Ad Hoc Netw.(2010), doi:10.1016/j.adhoc.2010
15