cross-layer design in cognitive radio networks
TRANSCRIPT
Some Cross-Layer Design and Performance Issues inCognitive Radio Networks
S.M. Shahrear Tanzil
M.A.Sc. Student
School of Engineering
The University of British Columbia Okanagan
Supervisor: Dr. Md. Jahangir Hossain
September 5, 2013
S.M. Shahrear Tanzil (UBC) 1 / 33 September 5, 2013 1 / 33
Outline
1 Introduction
2 Multi-Class Service Transmission over Cognitive Radio Network
3 Cross-Layer Performance in Presence of Sensing Errors
S.M. Shahrear Tanzil (UBC) 2 / 33 September 5, 2013 2 / 33
Introduction
Introduction
Figure 1: Fixed spectrum access policy
S.M. Shahrear Tanzil (UBC) 3 / 33 September 5, 2013 3 / 33
Introduction
Fixed Spectrum Access
A certain portion of radio spectrum is allocated/reserved for a certaingroup of users usually referred to as primary users (PUs)
Other group of potential users, usually referred to as secondary users(SUs) are not allowed to access the spectrum, even if a particularportion of the spectrum is currently not being used by the PUs
Recent studies on spectrum measurements have revealed that a largeportion of the assigned spectrum is used sporadically by the PUs
S.M. Shahrear Tanzil (UBC) 4 / 33 September 5, 2013 4 / 33
Introduction
Dynamic Spectrum Access
SUs can share the assigned spectrum with the PUs opportunistically
Underlay methodOverlay method
Spectrum hole Spectrum hole
Spectrum hole Spectrum hole
Spectrum hole
Fre
qu
en
cy
Time
Figure 2: An example of overlay spectrum access with spectrum holes
S.M. Shahrear Tanzil (UBC) 5 / 33 September 5, 2013 5 / 33
Introduction
Cognitive Radio
Facilitate dynamic spectrum access
Joseph Mitola proposed the concept of cognitive radio (CR)technology in 1998
Senses the spectrum of the PUs
Adapts various transmission and operating parameters including thefrequency range, modulation type, and power according to thewireless environment
S.M. Shahrear Tanzil (UBC) 6 / 33 September 5, 2013 6 / 33
Introduction
Motivations
Wireless channel quality not only varies with time but also theavailability of radio spectrum depends on PUs’ activity
Multi-class services e.g., video conferencing, email transfer and webbrowsing have diverse quality of service (QoS) requirements in termsof delay and packet loss probability
One design challenge: How to develop innovative resource allocationmechanisms that can meet diverse QoS requirements of differentclasses of services transmitted over the cognitive radio network (CRN)
Another design challenge: Channel sensing errors
S.M. Shahrear Tanzil (UBC) 7 / 33 September 5, 2013 7 / 33
Introduction
CRN Architecture: Infrastructure-Based
SU-1
CR BSPU BS
PU BS
SU-2
SU-K
Figure 3: Infrastructure-based CRN, CR=cognitive radio, BS=base station,PU=primary user, SU=secondary user.
S.M. Shahrear Tanzil (UBC) 8 / 33 September 5, 2013 8 / 33
Introduction
Operating Assumptions
PUs’ activity: ON/OFF
Channel: Slowly time varying, Nakagami-m, finite state Markovchannel
Channel scheduling: Max rate
S.M. Shahrear Tanzil (UBC) 9 / 33 September 5, 2013 9 / 33
Introduction
Cross-Layer Design
Physical layer
Data link layer
Pa
cke
ts fro
m h
igh
er
laye
r
Queue
Packet are transmitted
through the physical layer
Figure 4: Cross-layer design
Packet arrival follows batch Bernoulli random process
Packets are stored in the data link layer’s buffer/queue
Adaptive modulation and coding is employed
S.M. Shahrear Tanzil (UBC) 10 / 33 September 5, 2013 10 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Multi-Class Service
Voice, video streaming and web browsing have stringent delayconstraints i.e., delay sensitive (DS) service
Email has no stringent delay constraint i.e., delaynon-sensitive/best-effort (BE) service
S.M. Shahrear Tanzil (UBC) 11 / 33 September 5, 2013 11 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Rate Allocation Mechanism for a Particular SU
Buffer of delay sensitive service, Q(d)
Buffer of best effort service, Q(b)
Delay sensitive
packet arrival from
upper layer, β
Best effort
packet arrival
from upper
layer, α
kth user
rate
allocator
Allocated total transmission rate to user k
Rate of
delay
sensitive
service, R(d)
Rate of best
effort
service, R(b)
Figure 5: Rate allocation for multi-class service transmission for kth SU.
S.M. Shahrear Tanzil (UBC) 12 / 33 September 5, 2013 12 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Optimal Rate Allocation Mechanism
Formulated the problem as a constrained Markov decision process(MDP)
Objective
minimizex(S,A)
[d(d ,o)(S,A)]ᵀx(S,A) (1)
subject to:[p(b,o)loss (S,A)]ᵀx(S,A) ≤ p
(b)th (2)
[p(d ,o)loss (S,A)]ᵀx(S,A) ≤ p
(d)th (3)
p(b)th and p
(d)th are target packet loss probabilities of BE service and DS
service, respectively
S.M. Shahrear Tanzil (UBC) 13 / 33 September 5, 2013 13 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Optimal Rate Allocation Mechanism
x∗(S,A) denotes the probability of taking action A in state S thatminimizes the average queuing delay of DS packets while satisfiespacket loss probability constraints
From the optimal values, x∗(S,A) one can calculate QoS parameterse.g., packet loss probabilities and queuing delay
The optimal policies for constrained MDP are random
S.M. Shahrear Tanzil (UBC) 14 / 33 September 5, 2013 14 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Suboptimal Rate Allocation Mechanism
1: if Available transmission rate,R ≤ number of packets in the DS buffer then
2: R(d) ← R3: R(b) ← 04: else5: R(d) ← number of packets in the DS buffer6: R(b) ← R − R(d)
7: end if
S.M. Shahrear Tanzil (UBC) 15 / 33 September 5, 2013 15 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Suboptimal Rate Allocation Mechanism
Developed a queuing analytic model with the suboptimal rateallocation mechanism
Analyzed queuing analytic model as a quasi-birth-death (QBD)process
Calculated packet loss probabilities and queuing delay i.e., delaydistribution from the steady state probabilities of the QBD
S.M. Shahrear Tanzil (UBC) 16 / 33 September 5, 2013 16 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Numerical Results: Cumulative Distribution ofDelay of DS Packets
0 10 20 30 40 500
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
X: 10Y: 0.7156
time slots (X)
Pro
b.(d
elay
≤ X
) of
DS
pac
kets
X: 10Y: 0.8304
Suboptimal,K=2(ana)Optimal,K=2(sim)Suboptimal,K=3(ana)Optimal,K=3(sim)Suboptimal,K=4(ana)Optimal,K=4(sim)
Figure 6: Effect of number of SUs (K ) on the delay distribution of DS packets(ana=analysis, sim=simulation)
S.M. Shahrear Tanzil (UBC) 17 / 33 September 5, 2013 17 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Numerical Results: Packet Loss Probability of DSservice
1 2 3 4 50
0.01
0.02
0.03
0.04
0.05
0.06
X: 5Y: 0.05227
Number of secondary users
Pac
ket l
oss
prob
abili
ty o
f DS
ser
vice
X: 4Y: 0.03542
Suboptimal,(ana)
Optimal,(ana)
Figure 7: Effect of number of SUs (K ) on the packet loss probability of DSservice (ana=analysis, sim=simulation)
S.M. Shahrear Tanzil (UBC) 18 / 33 September 5, 2013 18 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Numerical Results: Packet Loss Probability of BEservice
1 2 3 4 50
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
X: 3Y: 0.06752
Number of secondary users
Pac
ket l
oss
prob
abili
ty o
f BE
ser
vice
X: 2Y: 0.008732 Suboptimal,(ana)
Optimal(ana)
Figure 8: Effect of number of SUs (K ) on the packet loss probability of BEservice (ana=analysis, sim=simulation)
S.M. Shahrear Tanzil (UBC) 19 / 33 September 5, 2013 19 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Application of the Developed Queuing Model with theSuboptimal Mechanism: Example
Table 1: Number of SUs for given QoS requirements (D(d,s)t,max = 10 (time slots)
with probability=0.8,P(d,s)t,loss ≤ 0.05 and P
(b,s)t,loss ≤ 0.05)
KD
(d,s)t,max
KP
(d,s)t,loss
KP
(b,s)t,loss
Ks
3 4 2 2
S.M. Shahrear Tanzil (UBC) 20 / 33 September 5, 2013 20 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Summary: Part I
Studied rate allocation mechanisms that allocate rate between twodifferent classes of services of a particular SU
Formulated the optimal rate allocation mechanism as a MDP
Also proposed a low-complexity suboptimal rate allocation mechanism
The performance of the suboptimal rate mechanism is quite similar tothe optimal rate allocation mechanism
Developed queuing analytic model with the suboptimal mechanism isuseful not only for calculating QoS parameters but also in making acall admission control decision
S.M. Shahrear Tanzil (UBC) 21 / 33 September 5, 2013 21 / 33
Multi-Class Service Transmission over Cognitive Radio Network
Publication
S M Shahrear Tanzil, Md. Jahangir Hossain, and Mohammad MRashid ,“Rate allocation mechanisms for multi-class service transmission overcognitive radio networks”, accepted inIEEE Global Commun. Conf. (Globecom’13), Atlanta, USA, Dec.2013.
S.M. Shahrear Tanzil (UBC) 22 / 33 September 5, 2013 22 / 33
Cross-Layer Performance in Presence of Sensing Errors
Sensing Errors in Cognitive Radio Systems
Two types of sensing errors i.e., false alarm and miss-detection
False alarm: CRN may detect a channel being used by PUs where inreality the channel is idle/PUs are not using the channel
Miss-detection: CRN may not be able to detect an active PU
S.M. Shahrear Tanzil (UBC) 23 / 33 September 5, 2013 23 / 33
Cross-Layer Performance in Presence of Sensing Errors
Random Transmission Protocol
Traditional deterministic protocol: If the channel is sensed as busy,CR base station (BS) decides to transmit with probability, 0
False alarm: Sensed as busy but in reality idle
Random transmission protocol: If the channel is sensed as busy, CRBS decides to transmit with probability, P1
S.M. Shahrear Tanzil (UBC) 24 / 33 September 5, 2013 24 / 33
Cross-Layer Performance in Presence of Sensing Errors
Random Transmission Protocol
Traditional deterministic protocol: If the channel is sensed as idle, CRBS decides to transmit with probability, 1
Miss-detection: Sensed as idle but in reality busy
Random transmission protocol: If the channel is sensed as idle, CRBS decides to transmit with probability, P2
S.M. Shahrear Tanzil (UBC) 25 / 33 September 5, 2013 25 / 33
Cross-Layer Performance in Presence of Sensing Errors
Queuing Analytic Model
Developed a queuing analytic model
Analyzed the queuing analytic model as a QBD process
Calculated QoS parameters of SUs e.g., packet loss probability andqueuing delay as well as QoS parameters of PUs e.g., collisionprobability from the steady state probabilities of the QBD
S.M. Shahrear Tanzil (UBC) 26 / 33 September 5, 2013 26 / 33
Cross-Layer Performance in Presence of Sensing Errors
Numerical Results: Packet Loss Probability
1 2 3 4 5 6 7 8 9 100.13
0.135
0.14
0.145
X: 8Y: 0.1401
Number of secondary users (K)
Pac
ket l
oss
prob
abili
ty
P1=0,P2=1(ana)P1=0.1,P2=1(ana)P1=0,P2=0.9(ana)
Figure 9: Effect of the values of P1, P2 and SUs (K ) on the packet lossprobability (ana=analysis).
S.M. Shahrear Tanzil (UBC) 27 / 33 September 5, 2013 27 / 33
Cross-Layer Performance in Presence of Sensing Errors
Numerical Results: Average Queuing Delay
1 2 3 4 5 6 7 8 9 1074.5
75
75.5
76
76.5
77
77.5
Number of secondary users (K)
Ave
rage
que
uein
g de
lay
(tim
e sl
ots)
P1=0,P2=1(ana)P1=0.1,P2=1(ana)P1=0,P2=0.9(ana)
Figure 10: Effect of the values of P1,P2 and SUs (K ) on average queuing delay(ana=analysis).
S.M. Shahrear Tanzil (UBC) 28 / 33 September 5, 2013 28 / 33
Cross-Layer Performance in Presence of Sensing Errors
Numerical Results: Collision Probability
1 2 3 4 5 6 7 8 9 100.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
X: 8Y: 0.07117
Number of secondary users (K)
Ove
rall
colli
sion
pro
babi
lity
X: 7Y: 0.06684
P1=0.1,P2=1(ana)P1=0,P2=0.9(ana)P1=0,P2=1(ana)
Figure 11: Effect of the values of P1, P2 and SUs (K ) on the collision probability(ana=analysis).
S.M. Shahrear Tanzil (UBC) 29 / 33 September 5, 2013 29 / 33
Cross-Layer Performance in Presence of Sensing Errors
Application of the Developed Model: Example
Table 2: Transmission Probabilities (P1,P2) vs. number of SUs for given QoSrequirements (pt,ploss = 0.14, Dt,avg = 77 (time slots) and pt,col = 0.07 )
(P1,P2) Kpt,ploss KDt,avg Kpt,col Ks
(0,0.9) 6 9 15 6(0,1) 7 11 13 7
(0.1,1) 8 12 8 8
S.M. Shahrear Tanzil (UBC) 30 / 33 September 5, 2013 30 / 33
Cross-Layer Performance in Presence of Sensing Errors
Summary: Part II
Investigated the performance of a random transmission protocol
Developed a queuing analytic model in presence of sensing errors
Calculated different QoS parameters using the developed queuingmodel
The queuing analytic model is also useful for admission control
Selected numerical results have shown that random transmissionprotocol can support more SUs than the classical deterministictransmission protocol
S.M. Shahrear Tanzil (UBC) 31 / 33 September 5, 2013 31 / 33
Cross-Layer Performance in Presence of Sensing Errors
Publication
S M Shahrear Tanzil and Md. Jahangir Hossain,“Cross-layer performance analysis for cognitive radio network with arandom transmission protocol in presence of sensing errors”,in Proc. of Int. Conf. on Cognitive Radio Oriented Wireless Networks(CROWNCOM’13), Washington DC, USA, Jul. 2013.
S.M. Shahrear Tanzil (UBC) 32 / 33 September 5, 2013 32 / 33