Performance Evaluation of Cognitive Radio Network Predictive MAC (P-MAC) Access Algorithm and its Enhancement Hanan Hussein, Hussein A. Elsayed, Salwa Elramly
Electronics and Communication Eng. Dept. Faculty of Engineering, Ain Shams University Cairo, Egypt
[email protected],
[email protected],
[email protected]
Cognitive radios, with their capabilities to sense the operating environment, learn and adapt in real time according to environment creating a form of mesh. Common Medium MAC protocols for a single channel do not provide in general, mechanisms for channel switching and are working with less performance in multiple channels environments. An enhanced Multichannel MAC protocol is needed when having multiple independent channels to be used simultaneously. Thus, several desired features are required for CR MAC protocol. First, it should be able to predict future spectrum usage based on statistics of local spectrum utilization up to the current time instance. To implement this feature, a CR device should monitor the spectrum usage continually to maintain an accurate view of spectrum utilization or depends on a statistical distribution for the current channel or have a database [4],[5] for certain bands such as TV bands. Second, it should avoid the harmful interference with licensed users. Third, it is preferred to have cooperation between the CR users to circumvent the collision between SUs. Here, we describe a prediction model for a proposed protocol called Predictive MAC (P-MAC) using Exponential Smoothing Model (ESM) [6]. This model works on an ON/OFF channel scenario with various distributions, e.g. Exponential distribution, and Pareto distribution. Exponential smoothing model is used to predict the channel distribution. In the ON/OFF channel scenario, the P-MAC exploits the OFF state duration for data transmission. This MAC protocol can support SUs in any CRN by using only one transceiver with the following advantages: (1) A Multi Channel Hidden Terminal Problem [7] has been solved (MCHTP). (2) Higher throughput for the OFF state than the other MAC protocols. (3) Working on distributed or centralized networks. (4) Less sensing overhead time. (5) Working on slotted and non slotted structure. The remaining of this paper is organized as follows: Section II provides a review of different MAC protocols and states their characteristics. In Section III, we introduce the P-MAC to access the available channel and illustrate its mechanism. Simulation results are introduced in Section IV while Section V concludes the paper.
Abstract—Cognitive radio (CR) technology is envisaged to solve the problems in wireless networks resulting from the limited available spectrum resources and the inefficiency in the spectrum usage by exploiting the existing wireless spectrum opportunistically. In this paper, Predictive MAC (P-MAC) is the MAC protocol that we propose as a new one for Cognitive Radio Networks (CRN). This MAC protocol can be applied on centralized or distributed networks. It is based on a prediction model called Exponential Smoothing model (ESM) [6]. The ESM predicts the channel distribution behavior for an ON/OFF channel scenario with various distributions. The PMAC has the ability to fill the channel’s vacancies as much as possible with less sensing time and RTS/CTS exchange. Therefore, the proposed model occupies the vacancies much better than the other ones. Also by using a simulation model we proved that it has higher throughput and less sensing overhead. Keywords-medium access control; spectrum cognitive radionetwork; exponential smoothing;
I.
sensing;
INTRODUCTION
The radio frequency spectrum is a scarce natural resource and its efficient use is of the greatest importance. The spectrum bands are usually licensed to certain services, e.g., mobile, fixed,broadcast, and satellite, to avoid harmful interference between different networks to affectusers. Most spectrum bands are allocated to certain services but worldwide spectrumoccupancy measurements show that only portions of the spectrum band are fully used. Moreover, there are large temporal and spatial variations in the spectrum occupancy [1]. In thedevelopment of future wireless systems the spectrum utilization functionalities will play a key role due to the scarcity of unallocated spectrum. Moreover, the trend in wirelesscommunication systems is going from fully centralized systems into the direction of self-organizing systems where individual nodes can instantaneously establish ad hoc networks whose structure is changing over time. Cognitive Radio Network (CRN) [2],[3] allows Secondary Users (SUs) or unlicensed users to share the licensed bands with the Primary Users (PUs) or the licensed users under some constraints such as limitation in the transmitted power.
978-1-4673-5742-5/13/$31.00 ©2013 IEEE
434
ICOIN 2013
II.
1) Sensing period: first it senses the channel to ensure the absence of PU. The SU will measure the energy level to detect the presence of PU, e.g. pilot carrier detection [5]. The time required in sensing each channel is called T sen. IEEE 802.22 imposes [4] that this sensing time be less than 1ms by using energy detection. In case of PU existence or there is another SU transmiting data,the SU will sense continuously till the channel becomes free.
RELATED WORK
As far as MAC protocol for Cognitive Radio Networks goes, research is still in its progress. The IEEE 802.22 working group is in the process of standardizing a centralized MAC protocol that enables SUs to reuse spectrum that operates on the TV broadcast bands [4].Two spectrum sensing techniques were proposed for this standard which are energy detection and feature detection, as in [5]. In [7]-[10] decentralized cognitive radio MAC protocols were proposed for coordinating spectrum access without harmful interference with the PU. These protocols were defined as Cognitive Radio Ad Hoc Networks (CRAHNs). An Energy efficient Cognitive Radio multichannel Medium Access Control (ECR-MAC) is a MAC protocol used for CRAHN in [7]. ECR-MAC analyzes an interference model which defines a set of links to the SU that can be active simultaneously without interfering. It also provides a solution for the MCHTP. A SYNchronized MAC protocol for multi-hop CRNs (SYN-MAC) is proposed in [8] where the usage of common control channel (CCC) is avoided. The scheme is applicable in heterogeneous environments where channels have different bandwidths and frequencies of operation. It states several disadvantages in using CCCH: the network may suffer from CCCH saturation problem and the usage of the CCCH wastes a spectrum resource which leads to reductionin the throughput of the CRN. SYN-MAC aims to solve both the CCCH problem and MCHTP, requiring CR users be attached with extra transceiver for channel sensing which degrades SUs performance and increases energy consumption. Reference [9] provided two periodic sensing schemes for SUs in cognitive radio network namely, periodic single sensing scheme and periodic continuous sensing scheme. The proposed schemes are applicable to any ON/OFF channel scenario. Opportunistic Periodic MAC protocol (OP-MAC) for CRN proposed in [10] is modeled based on renewal theory and applied on an ON/OFF channel scenario. This MAC protocol is based on SYN-MAC. By using one transceiver, OP-MAC solved the MCHTP, avoided the CCCH usage and provided better spectrum utilization. References [9], [10] proposed analysis based on single-SU single-channel scenario. All ON/OFF durations are assumed to be independent. This protocol assumes that SU transmits its data in the OFF state duration of the channel whenever it becomes available. III.
Sense Tsen Busy
Medium State
Free Set BO and sense
Busy
Medium State
Free Wait DIFS and sent RTS
Wait SIFS
CTS received?
No
Yes Start Transmission Figure1. The flow chart of Sensing and RTS/CTS exchange period [11]
THE PROPOSEDP-MAC
Here, we propose a detailed explanation for P-MAC protocol. Starting with negotiation and transmission phases, then we analyze the mathematical model which is used in the prediction.
Figure2. The MAC periods which exploit the OFF state durations [11]
A. P-MAC description Figure 1 and the following steps illustrate the required procedures to be executed each time the SU desires to transmit data to another SU:
2) RTS/CTS exchange period: after detecting the availability of this channel, SU negotiates with the other SUs to reserve this channel. Control Channel or Control slot
435
is used by transmitter to report other SUs with the reserved channel to solve the problem of multi channel hidden terminal. All CR users follow thestandard CSMA/CA protocol and Request To Send/Clear To Send (RTS/CTS) similar to 802.11 Distributed Coordination Function (DCF) mechanism [12] to access the control channel.To be more clear, in this period, before the sender reports the channel that it is going to access, it listens to the control channel and waits until it becomes idle then it sets a random backoff timer and senses during this period. After that, it waits for the control channel to remain idle for Distribution Inter Frame Space (DIFS) duration. Request To Send (RTS) is transmitted by the sender to the receiver after the medium is still idle. The RTS includes the source address, the destination address,Network Allocation Vector (NAV) and a list of channels the sender is willing to use. Upon a successful transmission of RTS, the receiver selects a channel and transmits a Clear To Send (CTS) to the transmitter after a Short Inter Frame Space (SIFS).The CTS contains the selected channel, the source address, the destination address and Network allocation Vector (NAV).Therefore,the total time required for this period is Tneg as obvious in (1).
where α is the smoothing factor, and 0 < α < 1. In OP-MAC [10], each time the SU desires to transmit data, it senses and negotiates during T0 before transmitting data in its MAC period Tp ,where T0 = Tsen + Tneg. Tp is specified as a fixed small period, independent of OFF state durations, e.g. 3ms, 5ms and 15ms as we stated before. As a result, transmitting data in (Tp – T0) every Tp by SU leads to reduction in the channel throughput and increase in the sensing overhead (Tsen-over = T0 / TP). In our proposed model, the transmission period is equal to (TMAC – T0) which is larger than the transmission period of OP-MAC (Tp – T0). Hence, we can raise the channel throughput and decrease the sensing overhead (Tsen-over = T0 / TMAC) with acceptable collision rate between primary and secondary user.
C. Enhanced P-MAC In this subsection, we present a modified version from PMAC protocol which is called Enhanced P-MAC (EPMAC). This protocol tries to fill the vacancies as much as possible more efficiently than P-MAC. After the SU finishes data transmission, it will sense the channel again. If the predicted MAC period TMAC is smaller than the OFF state time Toff then it will try to fill the remaining vacancies with small transmission periods, e.g. 3ms or 5ms, till the OFF Where TRTS is the required time for transmitting RTS, TCTS is state ends. This technique will raise the throughput more the required time for transmitting CTS and BO is the than P-MAC and OP-MAC but it will also raise the collision required time for the random backoff. By using this ratio. Some services can stand working with some mechanism, MCHTP can be solved [7]. interference level for a limited specific time so we can apply In case the sender fails in the negotiation process for any this technique on those services such as TV users which can stand interference within 2 sec. due to IEEE 802.22 [4],[5]. reason, e.g. collision happens with other SUs, it starts the We assume that always the SUs have data to transmit negotiation again till it succeeds in reaching to the receiver. whenever the OFF state becomes available. In the next 3) Transmission period: SU will transmit its data in this section, we will propose two scenarios. The first one is a period after sensing and RTS/CTS exchange period. The single transmitter and a single receiver (Single Flow). The transmission period will depend on the channel OFF state transmitter will exploit all the MAC periods Tp. While the duration to avoid the collision with PU. When starting this second scenario has multi-SUs (Multi-Flows) which will use period, the SU can't detect the appearance of the PU until it random access technique to access this MAC period. finishes transmitting its data as it has only one transceivr. In IEEE 802.22, it allows interference with PUs within IV. PERFORMANCE EVALUATION Channel Detection Time (CDT) [4],[5]. In this section, we present simulation results for single channel scenario employing ns2 (network-simulator2) VERSION NS-2.31.By applying Exponential and Pareto distributions, we evaluate three parameters; throughput, collision ratio, the validation of the prediction model.
B. Exponential smoothedP-MAC In P-MAC, we present a variable parameter called TMAC. TMAC is considered as the MAC period that the SU needs to access the channel's OFF state for sensing, negotiation and data transmission. TMAC is a predictive variable time based on the channel's OFF state durations (Toff).Figure 2 shows the predicted MAC periods which exist in the OFF state durations, the shaded region expresses the collision that may happen during transmission. These OFF durations vary randomly according to certain ON/OFF channel’s distribution, e.g. Exponential distribution [13], Pareto distribution [13]…etc. By applying Exponential Smoothed Model (ESM) [6], TMAC can be proposed as in (2) depending on the previous MAC period (TMAC-1) and the previous OFF state (Toff-1).
TABLE I contains the parameters that are used in the simulation results. In this table, it is notable that, we choose the data rate based on SU’s transceiver rate. In order to avoid the effect of this parameter, the results are presented in this section as a percentage or ratio instead of absolute value.
436
TABLE I parameters used in the proposed model
Parameter average ON time average OFF time Data Rate of PU Data Rate of SU Simulation Time CW T0 Tp Alpha (α)
TABLE II. Normalized throughput for the CR network.
Assigned value 5 sec 5 sec 200 Kbit/sec 100 Kbit/sec 100 sec 32 0.8 ms. 3 ms. 0.7
Exponential distribution Pareto distribution
P-MAC EP-MAC
EP-MAC 0.915149
0.656766
0.70148
0.880235
EP_MAC (RA)
EP_MAC (NRA)
P_MAC (RA)
P_MAC (NRA)
OP_MAC 1 0.8 0.6 0.4 0.2 0
1
2
4 8 15 Number of flows
20
B. Throughput performance of Multi Flows Here, normalized throughput is evaluated for the network through simulation with different number of flows at each time. In multi-users access, two techniques are introduced. The first technique is Random Access (RA) while the second technique is Not Random Access (NRA). NRA technique is based on selfishness of the users which means that the first transmitter which access the MAC period (TMAC) will transmit during the whole period. In RA technique, this period is shared for all SUs that desire to transmit data. In case of small number of users, NRA should have much higher throughput than RA as this requires less time for negotiation. However, this technique can’t achieve fairness between users. In the case of large number of transmitters, NRA has higher throughput because when collision happens, it will be in a small period. On the contrary, if collision happens in RA due to large number of SUs, the collision will be existed in the whole MAC period. Figure 4 shows the normalized throughput for OP-MAC, P-MAC and EP-MAC applying RA and NRA obtained from simulation results.
OP-MAC
100000 80000 60000 40000 20000 0
P-MAC 0.754826
Figure4. Normalized Throughput evaluation per number of flows for PMAC, OP-MAC and EP-MAC.
0.1 6.8 13.5 20.2 26.9 33.6 40.3 47 53.7 60.4 67.1 73.8 80.5 87.2 93.9
Network Throughput (bit/sec)
120000
Normalized throughput
A. Throughput performance of Single Flow Simulation results giving the throughput for the network in the case of P-MAC, OP-MAC and EP-MAC are compared in Fig3. Here we use a single SU which access the OFF state to show the performance of each protocol. Exponential distribution is used to evaluate the PU behavior. The throughput is measured only for the OFF states of the primary users. It is noteworthy from Fig. 3 that the throughput for P-MAC and EP-MAC are much higher than OP-MAC, as the later requires much higher negotiation than the other two protocols in case of long OFF states durations. Normalized throughput is defined as the ratio between the average throughput for the entire network and the traffic load. TABLE II shows this throughput for P-MAC, OPMAC and EP-MAC case of Exponential and Pareto distributions of PU ON/OFF. This table clarifies the performance analysis of each MAC protocol which proves that EP-MAC is the best of them.
OP-MAC 0.689917
Time (sec) Figure3. Throughput evaluation for P-MAC, OP-MAC and EP-MAC.
437
time increases, the collision ratio decreases due to (3). As was stated before, this ratio can be controlled in order not to exceed CDT [4],[5]. Also, the choice between P-MAC and EP-MAC can be based on the PU’s application, does it stand interference or not. From these tables, it is clear that the collision ratio for Pareto distribution is smaller than Exponential distribution. This happens because the standard deviation of Pareto distribution is larger than the standard deviation of exponential distribution. Therefore, for the same mean ON time for both distributions, we can find that the different ON states durations of the Pareto distribution is larger than the other distributions. In equation (3), if T on increases then the collision ratio decreases. It can also be noted that the collision ratio reduces for larger mean ON time which is an acceptable result.
0.95 0.9
Normalized Throughput
0.85 0.8 0.75 0.7 0.65 0.6 0.55
EP-MAC P-MAC OP-MAC
0.5 0.45
0
0.1
0.2
0.3
0.4
0.5 Alpha
0.6
0.7
0.8
0.9
1
Figure5. Normalized throughput evaluation versus α for P-MAC, OP-MAC and EP-MAC in case of single flow.
TABLE III Collision ratio calculation for OP-MAC, P-MAC and EP-MAC for mean ON time (5sec) and mean OFF time (5sec).
0.8 0.75 0.7
Normalized Throughput
OP-MAC 5% 3.9%
Exponential Dist. Pareto Dist.
P-MAC 17.9% 7.8%
EP-MAC 23.8% 14.8%
0.65
TABLE IV Collision ratio calculation for OP-MAC, P-MAC and EP-MAC for mean ON time (10 sec) and mean OFF time (10 sec).
0.6 0.55 0.5 0.45 EP-MAC-RA P-MAC-RA OP-MAC
0.4 0.35
0
0.1
0.2
0.3
0.4
0.5 Alpha
0.6
0.7
0.8
0.9
V.
P-MAC 9.7% 3%
EP-MAC 14.6% 5.2%
CONCLUSION
Cognitive radio networks might be one answer to overcoming spectrum scarcity. P-MAC has been proposed for cognitive radio ad hoc networks. It is applicable for any ON/OFF channel distribution. We use ESM to predict the OFF state duration of the PU to assign a suitable MAC period for SU’s data transmission.We also introduce an EPMAC protocol. These protocols proved that it can exploit the vacancies in the channel much better than OP-MAC with less sensing overhead ratio. Also with only one transceiver, they have higher average normalized throughput than other protocols with accepted collision ratio.
1
Figure6. Normalized throughput evaluation versus α for P-MAC, OP-MAC and EP-MAC in case of eight flows.
C. Prediction model validation Figures 5 and 6 show the performance of the prediction model using ESM in case of single flow and eight flows respectively. For different values of α, it is clear that this prediction model shows that the highest performance occurs at (α = 0.7). Therefore, we choose this value in our simulations.
ACKNOWLEDGMENT
D. Collision calculation Here, the collision ratio is defined as the overlapped area between the transmitted data of SU and the transmitted data of PU in the ON state duration, if it exists. This ratio is calculated in percentage as in (3).
OP-MAC 2.3% 1.9%
Exponential Dist. Pareto Dist.
The authors would like to thank Egyptian National Telecom Regulatory Authority (NTRA) for funding our project "Enhancement Proposals for DVB-T2 Systems and Cognitive Radio Networks Sharing the Same Frequency Band".
where Toff and Ton are the current OFF state and ON state durations, respectively. The percentage of collision ratio is shown in TABLE III and TABLE IV for different mean times. As the mean ON
REFERENCES [1]
[2]
438
Ian F. Akyildiz, Won-Yeol Lee, Mehmet C. Vuran, Shantidev Mohanty, "NeXt generation/dynamic spectrum access/cognitive radio wireless networks: A survey", in Computer Networks journal (Elsevier), Vol. 50, pp 2127–2159 , May 2006. Samuel Cheng, “Foundation of Cognitive Radio Systems,” March, 16, 2012, InTech.
[3] [4]
[5]
[6] [7]
[8]
[9]
Weng Chon Ao , “Connectivity of Multiple Cooperative Cognitive Radio Ad Hoc Networks,” IEEE Journal ON Selected Areas in Communications, Vol. 30, pp. 263-270, February 2012. Carlos Cordeiro, Kiran Challapali, Monisha Ghosh, “Cognitive PHY and MAC Layers for Dynamic Spectrum Access and Sharing of TV Bands,” in Proc. First Intl Workshop Technology and Policy for Accessing Spectrum (TAPAS06), New York, NY, USA, August 2006. Stephen J. Shellhammer, Ahmed K. Sadek and Wenyi Zhang, “Technical Challenges for Cognitive Radio in the TV White Space Spectrum,”Information Theory and Applications Workshop, in San Diego, CA, pp. 323-333, May 2009. Charle C. Holt, “Forecasting seasonals and trends by exponentially weighted moving averages,” in International Journal of Forecasting (Elsevier), Vol. 20, pp. 5-10, March 2004. S. M. Kamruzzaman, "An Energy Efficient Multichannel MAC Protocol for Cognitive Radio Ad Hoc Networks", International Journal of Communication Networks and Information Security (IJCNIS), Vol. 2, No. 2, pp. 112-119, August 2010. Y. R. Kondareddy and P. Agrawal, “Synchronized MAC Protocol For Multi-hop Cognitive Radio Networks”, IEEE International Conference on Communications (ICC 2008), Beijing, 2008.
[10]
[11]
[12] [13]
439
Dongyue Xue, Xinbing Wang, Ekram Hossain, “Optimization of Periodic Channel Sensing by Secondary Users in a Cognitive Radio Network”, GLOBECOM 2010, 2010 IEEE Global Telecommunications Conference, Miami, FL, pp. 1-5, Ref. Cited 13, Dec.2010. Dongyue Xue, Eylem Ekici and Xinbing Wang, “Opportunistic Periodic MAC Protocol for Cognitive Radio Networks,” GLOBECOM 2010 IEEE Global Telecommunications Conference, Miami, FL, pp. 1-6, Ref. Cited 13, Dec.2010. Hanan Hussein, Hussein A. Elsayed, Salwa Elramly, “A Predictive Model for MAC Protocol (P-MAC) of Cognitive Radio Network Using Exponential Smoothing Model,” MIC-WCMC 2012, MOSHARAKA International Conference, Palma de Mallorca, Spain, pp. 51-56, June 2012. G. Bianchi, “Performance Analysis of the 802.11 Distributed Coordination Funcition”, in IEEE Journal on Selected Areas in Communications,Vol. 18, No. 3, pp. 535-547, March 2000. Dan Rosbjerg, Henrik Madsen, Peter Funder Rasmussen, “Prediction in partial duration series with generalized pareto-distributed exceedances,” WATER RESOURCES RESEARCH, Vol. 28, No. 11, P. 3001, 1992. .
