Some Cross-Layer Design and Performance Issues in Cognitive Radio Networks

Transcription

1 Some Cross-Layer Design and Performance Issues in Cognitive 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, / 33

2 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, / 33

3 Introduction Introduction S.M. Shahrear Tanzil (UBC) 3 / 33 September 5, / 33

4 Introduction Fixed Spectrum Access A certain portion of radio spectrum is allocated/reserved for a certain group 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 particular portion of the spectrum is currently not being used by the PUs Recent studies on spectrum measurements have revealed that a large portion of the assigned spectrum is used sporadically by the PUs S.M. Shahrear Tanzil (UBC) 4 / 33 September 5, / 33

5 Introduction Dynamic Spectrum Access SUs can share the assigned spectrum with the PUs opportunistically Underlay method Overlay method Frequency Spectrum hole Spectrum hole Spectrum hole Spectrum hole Spectrum hole Time Figure 2: An example of overlay spectrum access with spectrum holes S.M. Shahrear Tanzil (UBC) 5 / 33 September 5, / 33

6 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 the frequency range, modulation type, and power according to the wireless environment S.M. Shahrear Tanzil (UBC) 6 / 33 September 5, / 33

7 Introduction Motivations Wireless channel quality not only varies with time but also the availability of radio spectrum depends on PUs activity Multi-class services e.g., video conferencing, transfer and web browsing have diverse quality of service (QoS) requirements in terms of delay and packet loss probability One design challenge: How to develop innovative resource allocation mechanisms that can meet diverse QoS requirements of different classes of services transmitted over the cognitive radio network (CRN) Another design challenge: Channel sensing errors S.M. Shahrear Tanzil (UBC) 7 / 33 September 5, / 33

8 Introduction CRN Architecture: Infrastructure-Based SU-1 PU BS CR BS SU-2 SU-K PU BS 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, / 33

9 Introduction Operating Assumptions PUs activity: ON/OFF Channel: Slowly time varying, Nakagami-m, finite state Markov channel Channel scheduling: Max rate S.M. Shahrear Tanzil (UBC) 9 / 33 September 5, / 33

10 Introduction Cross-Layer Design Data link layer Packets from higher layer Queue Physical layer 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, / 33

11 Multi-Class Service Transmission over Cognitive Radio Network Multi-Class Service Voice, video streaming and web browsing have stringent delay constraints i.e., delay sensitive (DS) service has no stringent delay constraint i.e., delay non-sensitive/best-effort (BE) service S.M. Shahrear Tanzil (UBC) 11 / 33 September 5, / 33

12 Multi-Class Service Transmission over Cognitive Radio Network Rate Allocation Mechanism for a Particular SU Delay sensitive packet arrival from upper layer, β Best effort packet arrival from upper layer, α Buffer of delay sensitive service, Q (d) Buffer of best effort service, Q (b) kth user rate allocator Rate of delay sensitive service, R (d) Rate of best effort service, R (b) Allocated total transmission rate to user k Figure 5: Rate allocation for multi-class service transmission for kth SU. S.M. Shahrear Tanzil (UBC) 12 / 33 September 5, / 33

13 Multi-Class Service Transmission over Cognitive Radio Network Optimal Rate Allocation Mechanism Formulated the problem as a constrained Markov decision process (MDP) Objective p (b) th and p(d) th minimize[d (d,o) (S, A)] x(s, A) (1) x(s,a) subject to:[p (b,o) loss (S, A)] x(s, A) p (b) th (2) [p (d,o) loss (S, A)] x(s, A) p (d) th (3) are target packet loss probabilities of BE service and DS service, respectively S.M. Shahrear Tanzil (UBC) 13 / 33 September 5, / 33

14 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 that minimizes the average queuing delay of DS packets while satisfies packet loss probability constraints From the optimal values, x (S, A) one can calculate QoS parameters e.g., packet loss probabilities and queuing delay The optimal policies for constrained MDP are random S.M. Shahrear Tanzil (UBC) 14 / 33 September 5, / 33

15 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) R 3: R (b) 0 4: else 5: R (d) number of packets in the DS buffer 6: R (b) R R (d) 7: end if S.M. Shahrear Tanzil (UBC) 15 / 33 September 5, / 33

16 Multi-Class Service Transmission over Cognitive Radio Network Suboptimal Rate Allocation Mechanism Developed a queuing analytic model with the suboptimal rate allocation mechanism Analyzed queuing analytic model as a quasi-birth-death (QBD) process Calculated packet loss probabilities and queuing delay i.e., delay distribution from the steady state probabilities of the QBD S.M. Shahrear Tanzil (UBC) 16 / 33 September 5, / 33

17 Multi-Class Service Transmission over Cognitive Radio Network Numerical Results: Cumulative Distribution of Delay of DS Packets X: 10 Y: Prob.(delay X) of DS packets X: 10 Y: Suboptimal,K=2(ana) Optimal,K=2(sim) 0.2 Suboptimal,K=3(ana) Optimal,K=3(sim) 0.1 Suboptimal,K=4(ana) Optimal,K=4(sim) time slots (X) 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, / 33

18 Multi-Class Service Transmission over Cognitive Radio Network Numerical Results: Packet Loss Probability of DS service 0.06 X: 5 Y: Packet loss probability of DS service X: 4 Y: Suboptimal,(ana) Optimal,(ana) Number of secondary users Figure 7: Effect of number of SUs (K) on the packet loss probability of DS service (ana=analysis, sim=simulation) S.M. Shahrear Tanzil (UBC) 18 / 33 September 5, / 33

19 Multi-Class Service Transmission over Cognitive Radio Network Numerical Results: Packet Loss Probability of BE service Packet loss probability of BE service X: 3 Y: X: Y: Suboptimal,(ana) Optimal(ana) Number of secondary users Figure 8: Effect of number of SUs (K) on the packet loss probability of BE service (ana=analysis, sim=simulation) S.M. Shahrear Tanzil (UBC) 19 / 33 September 5, / 33

20 Multi-Class Service Transmission over Cognitive Radio Network Application of the Developed Queuing Model with the Suboptimal 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) K D (d,s) t,max K P (d,s) t,loss K P (b,s) t,loss K s S.M. Shahrear Tanzil (UBC) 20 / 33 September 5, / 33

21 Multi-Class Service Transmission over Cognitive Radio Network Summary: Part I Studied rate allocation mechanisms that allocate rate between two different 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 to the optimal rate allocation mechanism Developed queuing analytic model with the suboptimal mechanism is useful not only for calculating QoS parameters but also in making a call admission control decision S.M. Shahrear Tanzil (UBC) 21 / 33 September 5, / 33

22 Multi-Class Service Transmission over Cognitive Radio Network Publication S M Shahrear Tanzil, Md. Jahangir Hossain, and Mohammad M Rashid, Rate allocation mechanisms for multi-class service transmission over cognitive radio networks, accepted in IEEE Global Commun. Conf. (Globecom 13), Atlanta, USA, Dec S.M. Shahrear Tanzil (UBC) 22 / 33 September 5, / 33

23 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 in reality 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, / 33

24 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, CR BS decides to transmit with probability, P 1 S.M. Shahrear Tanzil (UBC) 24 / 33 September 5, / 33

25 Cross-Layer Performance in Presence of Sensing Errors Random Transmission Protocol Traditional deterministic protocol: If the channel is sensed as idle, CR BS 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, CR BS decides to transmit with probability, P 2 S.M. Shahrear Tanzil (UBC) 25 / 33 September 5, / 33

26 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 and queuing delay as well as QoS parameters of PUs e.g., collision probability from the steady state probabilities of the QBD S.M. Shahrear Tanzil (UBC) 26 / 33 September 5, / 33

27 Cross-Layer Performance in Presence of Sensing Errors Numerical Results: Packet Loss Probability Packet loss probability X: 8 Y: P1=0,P2=1(ana) P1=0.1,P2=1(ana) P1=0,P2=0.9(ana) Number of secondary users (K) Figure 9: Effect of the values of P 1, P 2 and SUs (K) on the packet loss probability (ana=analysis). S.M. Shahrear Tanzil (UBC) 27 / 33 September 5, / 33

28 Cross-Layer Performance in Presence of Sensing Errors Numerical Results: Average Queuing Delay 77.5 Average queueing delay (time slots) P1=0,P2=1(ana) P1=0.1,P2=1(ana) P1=0,P2=0.9(ana) Number of secondary users (K) Figure 10: Effect of the values of P 1,P 2 and SUs (K) on average queuing delay (ana=analysis). S.M. Shahrear Tanzil (UBC) 28 / 33 September 5, / 33

29 Cross-Layer Performance in Presence of Sensing Errors Numerical Results: Collision Probability Overall collision probability X: 7 Y: X: 8 Y: P1=0.1,P2=1(ana) P1=0,P2=0.9(ana) P1=0,P2=1(ana) Number of secondary users (K) Figure 11: Effect of the values of P 1, P 2 and SUs (K) on the collision probability (ana=analysis). S.M. Shahrear Tanzil (UBC) 29 / 33 September 5, / 33

30 Cross-Layer Performance in Presence of Sensing Errors Application of the Developed Model: Example Table 2: Transmission Probabilities (P 1, P 2 ) vs. number of SUs for given QoS requirements (p t,ploss = 0.14, D t,avg = 77 (time slots) and p t,col = 0.07 ) (P 1, P 2 ) K pt,ploss K Dt,avg K pt,col K s (0,0.9) (0,1) (0.1,1) S.M. Shahrear Tanzil (UBC) 30 / 33 September 5, / 33

31 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 queuing model The queuing analytic model is also useful for admission control Selected numerical results have shown that random transmission protocol can support more SUs than the classical deterministic transmission protocol S.M. Shahrear Tanzil (UBC) 31 / 33 September 5, / 33

32 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 a random transmission protocol in presence of sensing errors, in Proc. of Int. Conf. on Cognitive Radio Oriented Wireless Networks (CROWNCOM 13), Washington DC, USA, Jul S.M. Shahrear Tanzil (UBC) 32 / 33 September 5, / 33

33 Cross-Layer Performance in Presence of Sensing Errors Thank you! S.M. Shahrear Tanzil (UBC) 33 / 33 September 5, / 33