Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 8, No. 5, October 2018, pp. 3767 3777 ISSN: 2088-8708 3767       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     Mark o vian Queueing Model f or Thr oughput Maximization in D2D-Enabled Cellular Netw orks Abiodun Gbenga-Ilori 1 and Olufunmilay o Sanusi 2 1 Department of Electrical and Electronics Engineering, Uni v ersity of Lagos, Lagos, Nigeria. 2 Computer and Electrical Engineering Department, Olabisi Onabanjo Uni v ersity , Ago-Iw o ye, Ogun State, Nigeria. Article Inf o Article history: Recei v ed July 21, 2017 Re vised May 19, 2018 Accepted June 21, 2018 K eyw ord: De vice-to-De vice (D2D) 5G cellular netw orks continuous-time Mark o v chain (CTMC) queueing model spectrum access ABSTRA CT De vice-to-De vice (D2D) communication has been considered a k e y enabling technol- ogy that can f acilitate spectrum sharing in 4 G and 5 G cellular netw orks. In order to meet the high data rate demands of these ne w generation cellular netw orks, this paper considers the optimization of a v ailable spectrum resource through dynamic spectrum access. The utilization of conti nuous-time Mark o v chain (CTMC) model for ef ficient spectrum access in D2D-enabled cellular netw orks is in v estig ated for the purpose of de- termining the impact of this model on the capacity impro v ement of cellular netw orks. The paper considers the use of CTMC model with both queueing and non-queueing cases called 13 -Q CTMC and 6 -NQ CTMC respecti v ely with the aim of impro ving the o v erall capacity of the c ellular netw ork under a f airness constrai nt among all users. The proposed strate gy consequently ensures that spectrum access for cellular and D2D users is optimally coordinated by des igning optimal spectrum access probabilities. Numerical simulations are performed to observ e the impact of the proposed Mark o vian queueing model on spectrum access and consequent ly on the capacity of D2D-enabled cellular netw orks. Results sho wed tha t the proposed 13 -Q CTMC pro vide a more spectrum- ef ficient sharing scheme, thereby enabling better netw ork performances and lar ger ca- pabilities to accommodate more users. Copyright c 2018 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Abiodun Gbeng a-Ilori Department of Electrical and Electronics Engineering Uni v ersity of Lagos, Lagos, Nigeria. gbeng ailori@unilag.edu.ng 1. INTR ODUCTION Mobile data traf fic, especially multimedia-rich services, are becoming a v ailable to more mobile users in recent years leading to an e v er -increasing demand for higher data rate wireless access. Examples of present netw orks that demand higher data rates are t h e Long T erm Ev olution-Adv anced (L TE-A) and W orldwide Inter - operability for Micro w a v e Access (W iMAX). There is also the ne xt generation 5 G netw ork which will require e v en higher data rates in order to pro vide services to users. Due to bandwidth limitation, it is vital to utilize tech- niques which can achie v e higher spectral ef ficienc y . T raditionally , the cellular netw ork operates on a centralized netw ork topology which is not spectral ef ficient since it requires that mobi le de vices communicate through the base station e v en when the y are in close proximity . As an alternati v e, D2D communication has been introduced to allo w peer -to-peer transmission among mobile de vices in close proximity , [1–3]. The adv antages of allo wing D2D communication underlay a cellular netw ork is that it can increase area spectral ef ficienc y , impro v e cellular co v erage, reduce l atenc y rate and also reduce ener gy consumption by mobile de vices [4]. Ho we v er , since D2D communi cation is lightly controlled by the base station, it poses a set of ne w challenges such as interference management and mode sel ection coordination. It is, therefore, necessary to ef ficiently and f airl y share the spectrum resource among cell ular users (CUs) and D2D users i n order to tak e full adv antage of the benefits of D2D communication and increase the o v erall capacity of the netw ork. A lot of research has been done in controlling interference in D2D communication underlaying cellular J ournal Homepage: http://iaescor e .com/journals/inde x.php/IJECE       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     DOI:  10.11591/ijece.v8i5.pp3767-3777 Evaluation Warning : The document was created with Spire.PDF for Python.
3768 ISSN: 2088-8708 netw orks, [5–11]. A fe w papers ha v e addressed this interference issue by controlling D2D access to the spectrum in a cellular netw ork [12–15]. V arious methods ha v e also been used in the past for the analysis and design of D2D spectrum sharing. In [16], the authors used a Poisson point process (PPP) to design a spectrum sharing mode for D2D-enabled cellular netw orks. In [17], in v estig ation of the throughput optimizat ion problem in D2D- underlaid cellular netw ork while prioritizing cellular services w as done. In [18], a mode selection algorithm to minimize outage probability and manage interference w as proposed. In [11], a technique for determining the minimum distance between simultaneously operating D2D links in order to determine the minimum required signal-to-interference-plus-noise ratio (SINR) at all recei v ers in the netw ork w as introduced. A similar method w as used in [7]. Some other papers used po wer control schemes for interference a v oidance in the netw orks, [7, 8, 13, 19–21]. Game theoretical approaches ha v e also been used to control interference and for ef ficient resource allocation, [5], [22–25]. Although the e xisting dynamic spectrum access schemes ha v e achie v ed some successes in enhancing spectrum ef ficienc y , most of them do not address f airness in heterogeneous netw orks, [21]. Besides maximizing the o v erall spectrum utilization, a good spectrum-sharing scheme should also be able to achie v e f airness among dissimilar users. The consequence of unf air resource allocation between dissimilar user s may result in spectrum resource w astage or redundant allocation, [26]. CTM C-based models ha v e been used before no w for analyzing the performance of cogniti v e radio netw orks (CRNs). Most importantly , it has been used to model the spectrum access of primary and secondary users in the CRN in order to achie v e an ef ficient, f air and fle xible spectrum sharing, [27–33]. In [32], an M/D/1 priority queueing scheme w as applied to e v aluate the performance of CRNs. In [33], a primary-prioritized Mark o v approach w as also used for dynamic spectrum access between secondary and primary users in CRNs. T o the best of our kno wledge, dynamic spectrum access schemes that can be used to impro v e the spectral ef ficienc y of D2D-enabled cellular netw orks has not been well in v estig ated. Moti v ated by the successes of CTMC models for ef ficient and f air spectrum sharing among dissimilar users in CRNs, this paper proposes an optimized spectrum access strate gy for combining CUs and D2D users in a cellular netw ork. CTMC model is use d with the aim of impro ving the o v erall capacity of the cellular netw ork under a f airness constraint among users. The proposed strate gy consequently ensures that there is no redundant allocation to a user while other users are in need of spectrum resource. Unlik e pre vious approaches, spectrum access for D2D users is optimally coordinated by designing optimal spectrum access probabilities. Consequently , Mark o vian queueing and non-queueing models are used for dynamic spectrum access where the cellular spectrum sub-band is shared by a CU and 2 D2D users and later e xtended to the analysis of a ge n e ral case with N D2D users. The quality of service (QoS) constraint is defined by an SINR threshold that the CU should absolutely achie v e. Hence depending on the channel state information recei v ed, the CU, D2D users or all N + 1 users can transmit on the same frequenc y band. The computation time of this comple x CTMC model consisting of one CU and N D2D users is also quite lo w . The k e y contrib utions of this paper can be summarized as follo ws: formulation of ef ficient spectrum access 6 -NQ CTMC and 13 -Q CTMC models to sho w t h e throughput g ain possible in D2D-enabled cellular netw orks, proposal of a 13 -Q CTMC model that ensures ef ficient and optimal spectrum access scheme for D2D users while protecting cellular users from intolerably high interference from D2D users, proposal of a 13 -Q CTMC model that reduces the connection set-up time and thereby reducing the o v erall latenc y in the cellular netw ork. The remaining part of the paper is or g anized as follo ws. Section 2 presents the system model and as - sumptions. Section 3 presents the proposed Mark o vian non-queueing model and computation of the probability of co-transmission for multiple D2Ds and CU ha ving SINR constraint. In Section 4, the Mark o vian queue- ing model is presented while the simulation studies are pro vided in Section 5. Finally , concluding remarks are pro vided in Section 6. 2. SYSTEM MODEL In this paper , a dynamic spectrum access model is used in a cellular netw ork where multiple D2D users are allo wed to underlay licensed CUs. A netw ork consisting of N D2D links and 1 CU de vice with licensed sub-band is considered. A sub-band is a frequenc y spectrum sub-allocated to a licensed cellular user . A cellular user o wns a licensed sub-band which it can share with a number of D2D links. Figure 1 sho ws the system diagram. The CU communicates solely t hrough the base station using link l while the tw o sets of D2D users communicate directly without the base station using links D 1 and D 2 . It is IJECE V ol. 8, No. 5, October 2018: 3767 3777 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE I SSN: 2088-8708 3769 Figure 1. D2D-Enabled Cellular Netw ork assumed that all communication occurs in the same cell within the same channel. The paper also assumed that the D2D links share the uplink resource with the cellular user equipment. D2D communication is allo wed as long as it does not cause SINR of the cellular link to drop belo w the required minimum. The SINR of the cellular link tak es a higher priority . The paper aims at determining the optimal spectrum access probabilities for each D2D link in the cellular netw ork. If optimal coordination of D2D spectrum access can be guaranteed, then it is possible to achie v e a good trade-of f between spect rum ef ficienc y and interference reduction. First, the spectrum access is modelled as a CTMC without queueing. In this case, if a D2D link does not meet the minimum requirement to underlay a cellular sub-band, it is dropped and has to start the process of spectrum search all o v er ag ain. The disadv antage of this is that the D2D users spend more time and battery po wer is lost while searching for ne w spectrum. The spectrum access is later modelled as a CTMC with queueing where D2D links that do not meet present spectrum use requirements h a v e the opportunity to queue up for a future time to access the spectrum instead of being dropped. This hopefully impro v es the netw ork throughput while reducing communication set- up time and battery po wer consumption. The non-queueing CTMC spectrum access is modelled as a si x-state CTMC while the queueing CTMC spectrum access is modelled as a thirteen-state CTMC. The se state diagrams are used to compute the spectrum access probabilities ( ) of being in each state. The a v erage throughput U for each user in the cellular netw ork is therefore computed as: U = c r c + d 1 r d 1 + + d n r d n ; (1) where the set = f c ; d 1 ; ; d n g is the spectrum access probabilities of the cellular user C and set of N D2D users D = [ d 1 ; d 2 ; ; d n ] and U is a function of and the channel capacity of each user r = f r c ; r d 1 ; ; r d n g . The channel capacity for a user A operating in the spectrum band alone is r 1 A = W l og 2 (1 + P A G AA n 0 ) ; (2) and the channel capacity for user A when it coe xists with another user B in the same spectrum band is r 2 A = W l og 2 (1 + P A G AA n 0 + P A 6 = B P B G B A ) ; (3) where W is the communication bandwidth, n 0 is the thermal noise po wer , P A and P B are the transmission po wer for users A and B respecti v ely , and G AA is the channel g ain for user A while G B A is the channel g ain from user B’ s transmitter to user A s recei v er . Using the channel state information (CSI) g athered, the base station e v aluates the spectrum utilization, computes the opti mal access probabilities in dif ferent states and sends the results to the D2D link. The queueing model helps to determine the w aiting period, if necessary , for each D2D link. Of course, if the w aiting period is unacceptably long, the D2D link may choose to use another cellular link or e v en an unlicensed band. If a D2D user is transmitting and a CU arri v es requesting the use of the channel and the minimum requirement is not met, the CU queue up and w ait for the D2D user to complete transmission. Ho we v er , if on arri v al of the CU, there are D2D users on the queue for spectrum use, the CU tak es priority o v er the D2D users in the queue. Mark o vian Queueing Model for Thr oughput Maximization in D2D-Enabled ... (Abiodun Gbenga-Ilori) Evaluation Warning : The document was created with Spire.PDF for Python.
3770 ISSN: 2088-8708 T able 1. The Six States of the 6 -NQ CTMC. S tate D escr ipti on 0 No user in the spectrum C CU in the spectrum D One D2D user in the spectrum 1 CU and one D2D user in the spectrum 2 Both D2D users in the spectrum 3 All users in the spectrum 3. CELLULAR-PRIORITIZED NON-Q UEUEING CTMC In this section, the dynamics of the system consisting of a CU and tw o D2D users is first modell ed using a CTMC without queueing and later generalized to multiple D2D users. The probabilities in v olv ed in these transitions are also computed and used to deri v e the throughput that can be achie v ed in the cellular netw ork. 3.1. 6 -NQ CTMC In this section, it is assumed that when a D2D user requesting spectrum access appears, the base stati on determines if the D2D meet the minimum spectrum access requirements needed in the cellular netw ork using the CSI. Otherwise, the D2D user is dropped and can either w ait for a later time to try ag ain, request for another cellular band or use an unlicensed spectrum band. The scenario is therefore modelle d as a six-state non-queueing CTMC. First, it is assumed that a maximum of three users can use t he single uplink frequenc y channel of the cellular user; 1 CU and 2 D2D users. The paper later e xtends to a more general case of N -D2D users. In the non-queueing CTMC model, the CU’ s priority , in terms of spectrum access, is not so ob vious. Ho we v er , the base station gi v es the CU higher data rates compared to the D2D users. The spectrum access of the cellular and D2D users are modelled as independent Poisson process with arri v al rates c and d respecti v ely . The servi ce times are assumed to be e xponentially distrib uted with departure rates for cellular and D2D users denoted as c and d respecti v ely . The six states of the non-queueing CTMC are described in T able 1. This six-state Mark o v chain is denoted by 6 -NQ CTMC for short. The spectrum access process is sho wn in Figure 2 . Assume at first that cellular band is idle, in which case t he 6 -NQ CTMC is in state 0 . In this case, there can be either an arri v al of a cellular user C or a D2D user d . If an y of these 2 users arri v e, the 6 -NQ CTMC transit to either state C or D with transition rates c and d respecti v ely . If user C or d complete service before an y other user requests spectrum access, 6 -NQ CTMC then transits to state 0 with departure rate c and d accordingly . Ho we v er , if a second D2D arri v es while the CU or the first D2D are in the spectrum, the 6 -NQ CTMC transits to either 1 or 2 accordingly with rate d . Once the CU or one of the D2D complete transmission, there is a transition to either state C or D with the departure rate of d . If both the CU and a D2D are in the spectrum and the second D2D requests a spectrum band, then the 6 -NQ CTMC can either transit to state 3 from 2 with an arri v al rate of c or transit to state 3 from state 1 with an arri v al rate of d . In all of the transitions described abo v e, it has been assumed that no tw o D2D users can arri v e or depart at e xactly the same time. This assumption is justified for independent Poisson processes. 0 D C 1 2 3 d c d c d d c d c d c d c d Figure 2. The Rate Diagram of 6-NQ CTMC IJECE V ol. 8, No. 5, October 2018: 3767 3777 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE I SSN: 2088-8708 3771 The CTMC model is represented by ( , Q ) where = f 1 ; 2 ; ; n g is the state space while Q is the transition rate matrix and in our 6 -NQ CTMC, = f 0 ; C ; D ; 1 ; 2 ; 3 g . Q = [ q ij ] ; (4) where Q in our 6 -NQ CTMC is gi v en as the matrix: Q = 2 6 6 6 6 6 4 ( c + d ) c d 0 0 0 c ( c + d ) 0 d 0 0 d 0 ( d + c + d ) c d 0 0 d c ( d + c + d ) 0 d 0 0 d 0 ( d + c ) c 0 0 0 d c ( d + c ) 3 7 7 7 7 7 5 From the matrix sho wn abo v e, q ii = P j 6 = i q ij and 0 q ij < 1 8 i 6 = j . The balance equation to be solv ed is Q = 0 and P n = 1 . Therefore the analysis of Figure 2 consists of the follo wing system of equations; 8 > > > > > > > > < > > > > > > > > : 0 ( c + d ) = C c + D d ; C ( c + d ) = 0 c + 1 d ; D ( d + c + d ) = 0 d + 1 c + 2 d ; 1 ( d + c + d ) = C d + D c + 3 d ; 2 ( d + c ) = D d + 3 c ; 3 ( c + d ) = 1 d + 2 c ; (5) 0 + C + D + 1 + 2 + 3 = 1 : (6) Equation (5) represents the flo w-balance at each of the six states and equation (6) represents the nor - malization equation that should satisfy a Mark o v chain with n being the steady state probabilities of being in a particular place where n 2 f 0 ; C ; D ; 1 ; 2 ; 3 g . From these equations, the a v erage throughput for the cellular ( U c ) and each of the tw o D2D users U d 1 and U d 2 can be deduced as follo ws: U c = C r c + 1 r 1 + 3 r 3 ; (7) U d 1 = U d 2 = D r d + 1 r 1 + 2 r 2 + 3 r 3 : (8) T otal a v erage throughput is therefore U = U c + U d 1 + U d 2 : (9) 3.2. Generalized-NQ CTMC Our CTMC can be generalized to model the scenario with 1 CU and N D2D users as sho wn in the rate diagram of Figure 3 . In this case, the state space S N Q has 2( N + 1) states. S N Q consists of a combination of the status of the CU and the N D2D users and can be written as: ( s N Q C U ; s N Q D 2 D ) 2 S N Q , ( idl e C U ; idl e D 2 D ) [ ( C U ; idl e D 2 D ) [ f ( C U ; D 2 D ) g [ f ( idl e C U ; D 2 D ) g ; (10) where ( idl e C U ; idl e D 2 D ) is a state (0 ; [0 ; ; 0]) , in which there is no user re qu e sting the spectrum. ( C U ; idl e D 2 D ) is a state (1 ; [0 ; ; 0]) in which only the CU is using the spectrum. The set of states f ( C U ; D 2 D ) g represent all the states where a combination of 1 CU and one or up to N D2D users are in the spectrum. The set of states f ( idl e C U ; D 2 D ) g represent all the states where there is no CU b ut one or up to N D2D users are in the spectrum. If q ij , f s i ! s j g denotes the transition from state s i to state s j , then we can construct the matrix Q = [ q ij ] . F or the state space S N Q = [ n 0 ; n 1 ; ; n g ; ; n N +1 ] where N + 1 is the number of users in the spectrum; CU or (and) D2D users, the number of transition states is gi v en as n = 2( N + 1) for N D2D users. Therefore, q f [ n 0 ; n 1 ; ; n g ; ; n N +1 ] ! [ n 0 ; n 1 ; ; 1 n g ; ; n N +1 ] g = g . W e can also solv e the stationary probability: s n = [ s 1 ; ; s (2( N +1)) ] using Q = 0 and P 2( N +1) n =1 s n = 1 . W e can re wri te this as: Q T 1 1 (2( N +1)) T = 0 (2( N +1)) 1 1 : (11) Mark o vian Queueing Model for Thr oughput Maximization in D2D-Enabled ... (Abiodun Gbenga-Ilori) Evaluation Warning : The document was created with Spire.PDF for Python.
3772 ISSN: 2088-8708 0 C D 0 D 1 C D 1 D N 1 C D ( N 1) D N C D N c d 1 c d 1 d 2 d 1 c d 1 d 2 c d 2 d N 1 d 2 d N 1 d N 1 d N d N 1 d N d N c d N c Figure 3. The Rate Diagram of the Generalized-NQ CTMC 4. CELLULAR-PRIORITIZED Q UEUEING CTMC In the pre vious section, D2D link that is unable to meet the QoS requirement of the netw ork is dropped by the base station. This means that the D2D link is compelled to seek alternati v e means of communication. This has some disadv antages especially with respect to ef ficient usage of spectrum resources and delays in the netw ork. In order to maximize the ef ficient use of these cellular channels, a concept w as introduced where spectrum requests by D2D links that do not meet QoS of the netw ork are queued in a b uf fer at the base station and the spectrum is immediately made a v ailable to D2D links on the queue without an y time lapse in the usage of the licensed cellular bands. This w ay the o v erall communi cation set-up time for de vices in the netw ork is greatly reduced and the D2D user can conserv e battery ener gy . First, this is modelled as a thirteen-state CTMC with queueing kno wn as 13-Q CTMC and then later generalized in what is called the Generalized-Q CTMC. 4.1. 13 -Q CTMC In this sub-section, the netw ork is modelled to depict a situation where a D2D link request the use of a cellular spectrum and the base station, using the CSI, determines if the D2D link meets the minimum QoS requirement of the netw ork. If it does, the D2D link is admitted into the channel. Ho we v er , if it does not meet this requirement, it is admitted into a queue and can access the spectrum at a later time. The thi rteen states of the 13-Q CTMC are described in T able 2 and the rate diagram is gi v en in Figure 4 . Then the equation array go v erning the abo v e system is gi v en by: 8 > > > > > > > > > > > > > > > > > > > > > > > > > < > > > > > > > > > > > > > > > > > > > > > > > > > : 0 ( c + d ) = 1 c + 2 d ; 1 ( c + 2 d ) = 0 c + 3 d + 6 d ; 2 ( d + c + 2 d ) = 0 d + 4 d + 8 c + 12 d ; 3 ( d + 2 d ) = 1 d + 5 d + 11 d ; 4 ( d + 2 c ) = 2 d + 5 c + 9 c ; 5 ( d + c ) = 3 d + 4 c ; 6 ( d + d ) = 1 d + 7 c ; 7 ( d ) = 6 d ; 8 ( c ) = 2 c ; 9 ( c ) = 2 c ; 10 ( c ) = 12 c ; 11 ( d ) = 3 d ; 12 ( d + c ) = 2 d + 10 c ; (12) 0 + 1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9 + 10 + 11 + 12 = 1 : (13) Ag ain the stationary probabilities can be solv ed by using Q = 0 and P n = 1 as sho wn in equations (12) and (13), and the total a v erage throughput in the netw ork can be determined from these equations. IJECE V ol. 8, No. 5, October 2018: 3767 3777 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE I SSN: 2088-8708 3773 T able 2. The Thirteen States of the 13 -Q CTMC. S tate D escr ipti on 0 (0 ; 0) No user in the spectrum 1 ( C ; 0) CU in the spectrum 2 ( D ; 0) 1 D2D user in the spectrum 3 (1 ; 0) CU and 1 D2D user in the spectrum 4 (2 ; 0) Both D2D users in the spectrum 5 (3 ; 0) All 3 users in the spectrum 6 ( C ; D w ) CU in the spectrum, 1 D2D user w aiting 7 ( C ; 2) CU in the spectrum, 2 D2D users w aiting 8 ( D ; C w ) 1 D2D user in the spectrum, CU w aiting 9 (2 ; C w ) Both D2D users in the spectrum, CU w aiting 10 ( D ; 1 w ) 1 D2D user in the spectrum, 1 D2D and CU w aiting 11 ( D ; D w ) 1 D2D user in the spectrum, 1 D2D w aiting 12 (1 ; D w ) CU and 1 D2D user in the spectrum, 1 D2D w aiting 4.2. Generalized-Q CTMC W e can generalize the CTMC with queueing model as: ( s Q C U ; s Q D 2 D ) 2 S Q , S N Q [ S W ; (14) where S Q is the state space for the queueing model, S N Q is as gi v en in equation (10). The state space for the w aiting incorporated is denoted by S W and it is gi v en as: S W = [ f ( in 1 user ; W 1 user ) g [ f ( in 1 user ; W 2 user s ) g [ [ f ( in 1 user ; W k 1 user s ) g [ [ f ( in 2 user s ; W k 2 user s ) g [ [ f ( in k 1 user s ; W 1 user ) g ] ; (15) where k is the number of users in the spectrum; CU and D2D users inclusi v e. ( in 1 user ; W k 1 user s ) means that 1 user is occup ying the spectrum, CU or D2D, and the other ( k 1) users are w aiting in the queue to use the spectrum. The generator matrix Q = [ q ij ] is ag ain constructed for s n = [ n 0 ; n 1 ; ; n g ; ; n k ] where n = N 2 + 4 N + 1 for N D2D users. q f s n ! s w n g denote transition that oc curs when a user j arri v es gi v en that the CSI does not support spectrum sharing with user j at that time. The transition goes to q f s w n ! s n g with the departure of some users and the accommodation of user j . The follo wing equation array is then solv ed to obtain our stationary probabilities: Q T 1 1 ( N 2 +4 N +1) T = 0 ( N 2 +4 N +1) 1 1 : (16) 5. NUMERICAL RESUL TS In this section, the system performance of the 6 -NQ CTMC and 13 -Q CTMC spectrum access schemes are e v aluated and analyzed in terms of the total throughput that can be achie v ed using each of them. The paper simulates a system with a cellular user and multiple D2D users arri ving according to Poisson process with arri v al rates c and d respecti v ely . MA TLAB is used to conduct the simulation e xperiments in order to determine the throughput in a cell using each of the tw o schemes discussed in sections III and IV . The goal is to compare the performance of the 6 -NQ CTMC model with that of the 13 -Q CTMC model in order to sho w the better performance of the proposed Mark o vian queueing model. The follo wing parameters were used: channel bandwidth = 5 M H z , UE transmitter po wer = 24 dB m , thermal noise per MHz = 114 dB , recei v er g ain = 0 dB i , c = 1 20 s 1 , d = 1 20 s 1 , c = 20 s 1 , d = 25 s 1 . The base station is located at the center of the 300 m radius cell and the CU and D2D users are distrib uted randomly around it. D2D links ha v e a maximum distance of 20 m . Mark o vian Queueing Model for Thr oughput Maximization in D2D-Enabled ... (Abiodun Gbenga-Ilori) Evaluation Warning : The document was created with Spire.PDF for Python.
3774 ISSN: 2088-8708 2 ; C w 2 ; 0 3 ; 0 1 ; 0 1 ; D w D ; 0 D ; D w D ; 1 w D ; C w 0 ; 0 C ; 0 C ; D w C ; 2 w c c d c c d d d d d d c d d c d c c c d d c d d d d Figure 4. The Rate Diagram of 13-Q CTMC Figure 5. A v erage Throughput Achie v able in Cell for 6 -NQ CTMC and 13 -Q CTMC Figure 6. Optimal Access Probability for CU and D2D Users The simulation is also used to determine the a v erage w aiting time in the entire system for both models. It is also used to determine the w aiting time in the queue for the 13 -Q CTMC model. This part of the simulation pro vides a w ay to assess the connection set-up time a nd o v erall latenc y in the system usi ng the proposed queueing model. F or determination of the w aiting time in the systems and queue, the con v entional method discussed in [34] is adopted. The performance of the proposed 13 -Q CTMC model is also v alidated by comparing it with the e xisting non-persistent carrier sense multiple access (CSMA) spectrum access model reported in [35] to further sho w its superiority . Figure 5 compares the throughput achi e v able by the proposed 13 -Q CTMC model with that of the 6 -NQ CTMC model and the e xisting non-persistent CSMA spectrum access technique. It can be sho wn, by comparing throughput achie v able from the three spectrum access schemes, that 13 -Q CTMC has the highest throughput while CSMA has the lo west throughput. The poor performance of CSMA is due to the collision rate and inef fic ient random w aiting time of this scheme. Ho we v er , by controlling the access probabilities of D2D users in both CTMC schemes, it w as possible to accommodat e more traf fic and greatly increase the throughput. The 13 -Q CTMC model w as able to achie v e further increase in throughput becaus e queueing UEs, instead of rejecting requests, impro v ed the o v erall throughput of the netw ork. Generally , there is a throughput de gradation for all models as increases. This is as a result of interference that may occur in the netw ork. Though there is a general de gradation in throughput as increases, yet a slo wer rate of de gradation w as noticed in the proposed 13 -Q CTMC model with a de gradation of 0 : 58% in throughput when w as increased from 1 to 4 and 5 : 53% de gradation in throughput when w as increased from 1 to 20 . In the 6 -NQ CTMC model, when w as increased from 1 to 4 , the netw ork e xperienced a 1 : 9% de gradation in throughput and when w as increased from 1 to 20 , 11 : 43% de gradation in throughput w as e xperienced in the netw ork. The CSMA performed v ery poorly with de gradation of 10 : 7% as increased from 1 to 4 and de gradat ion of 65% as increased from 1 to 20 . Therefore the 13 -Q CTMC w as able to achie v e the highest throughput and also has the best access scheme because it performed best with an increase in arri v al rate. IJECE V ol. 8, No. 5, October 2018: 3767 3777 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE I SSN: 2088-8708 3775 Figure 7. W aiting T ime in the System for 6 -NQ CTMC and 13 -Q CTMC Figure 8. W aiting T ime on the Queue for 13 -Q CTMC Figure 6 sho ws the a v erage access probability of D2D and cellular users. The access probability of the cellular user decreases as arri v al r ate of cellular users increases. The arri v al rate of the cellular user also causes a reduced access probability f o r the D2D user . Ho we v er , the access probability of the cellular user is still more than that of the D2D user because of the priority gi v en to the cellular user since its arri v al rate is fix ed at 20 s 1 . The service time in the netw ork for our 6 -NQ CTMC w as also compared with that of the 13 -Q CTMC. It can be sho wn from Figure 7 that the w aiting time in both models is comparably close when the number of arri v als, , is less than 4 users per seconds with 13 -Q CTMC model slightly better than the 6 -NQ CTMC model. Ho we v er , as the number of arri v als per second increases, the w aiting time in the system increases e xponentially in the 6 -NQ CTMC model. It is seen that the 13 -Q CTMC w as able to perform better in that it is able to toler ate interference in the system despite the inclusion of queueing time in this model. Figure 8 sho ws the w aiting time on the queue for the 13 -Q CTMC. W e see that e v en with the queue, users will still ha v e access to the system f aster than using the 6 -NQ CTMC. The proposed 13 -Q CTMC, therefore, sho wed better performances and lar ger capabilities to accommodate more users. It also of fers a more ef ficient spectrum utilization compared to 6 -NQ CTMC. 6. CONCLUSION In the paper , a Mark o vian-queueing approach is proposed for optimizing the use of cellular spectrum resource through dynamic spectrum access. This is highly necessary in order to meet the high data rate demands of 4 G and 5 G cellular netw orks. The use of a Mark o vian-queueing model kno wn as 13 -Q CTMC is proposed for underlaying D2D users in a cellular bandwidth in order to optimize the use of spectrum resource and increase throughput in the netw ork. A Mark o vian-queueing model is chosen because of the successes of CTMC models in achie ving ef ficient and f air spectrum sharing in heterogeneous netw orks. The proposed 13 -Q CTMC queueing model is compared with the 6 -NQ CTMC model that does not support queueing and the e xisting non-persistent CSMA spectrum access scheme. Simulation results sho wed that the proposed Mark o vian-queueing model is more ef ficient in the use of limited spectrum resource and also yielded better throughput in the netw ork compared to the other tw o spectrum access techniques. The 13 -Q CTMC model ensures ef ficient and optimal spectrum access scheme for D2D users while protecting cellular us ers from intolerably high interference from D2D users. Compared with the other spectrum access techniques, the 13 -Q CTMC model sho wed a considerable reduction in the connection set-up time and thereby reducing the o v erall latenc y in the cellular netw ork. A CKNO WLEDGMENT The first author will lik e to ackno wledge the support of the Ale xander v on Humboldt F oundation for financing her post-doctoral research stay at the Ruhr -Uni v ersit ¨ at Bochum, German y . REFERENCES [1] A. Krishna, A. Chakra v arth y , and A. Sastry , “A Hybrid Cryptographic System for Secured De vice t o De vice Communication, International J ournal of Electrical and Computer Engineering (IJECE) , v ol. 6, no. 6, pp. 2962–2970, December 2016. Mark o vian Queueing Model for Thr oughput Maximization in D2D-Enabled ... (Abiodun Gbenga-Ilori) Evaluation Warning : The document was created with Spire.PDF for Python.
3776 ISSN: 2088-8708 [2] A. Asadi, Q. W ang, and V . Mancuso, “A Surv e y on De vice-to-De vice Communic ation in Cellular Net- w orks, IEEE Communications Surve ys and T utorials , v ol. 2014, no. 4, pp. 1801–1819, April 2014. [3] A. Asadi and V . Mancuso, “Netw ork-assisted Outband D2D-clustering in 5G Cellular Netw orks: Theory and Practice, IEEE T r ansactions on Mobile Computing , v ol. PP , no. 99, pp. 1–15, October 2016. [4] L. Song, D. Niyato, Z. Han, and E. Hossain, “Game-Theoretic Resource Allocation Methods for De vice- to-De vice Communication, IEEE W ir eless Communication , v ol. 21, no. 3, pp. 136–144, June 2014. [5] Y . Xiao, K. Chen, C. Y uen, and L. DaSilv a, “Spectrum Sharing for De vice-to-De vice Communications in Cellular Netw orks: A Game Theoretic Approach, in Pr oc. IEEE Confer ence on Dynamic Spectrum Access Networks (D YSP AN) , McLean, V A, USA, 2014. [6] D. Feng, L. Lu, Y . Y uan-W u, G. Y . Li, G. Feng, and S. Li, “De vice-to-De vice Communications in Cellular Netw orks, IEEE Communications Ma gazine , v ol. 52, no. 4, pp. 49–55, April 2014. [7] D. Feng, L. Lu, Y . Y uan-W u, G. Y . Li, S. Li, and G. Feng, “De vice-to-De vice Communications Underlaying Cellular Netw orks, IEEE T r ans. on Communications , v ol. 61, no. 8, pp. 3541–3551, Aug. 2013. [8] P . J ¨ anis, C. Y u, K. Doppler , C. Ribeiro, C. W ijting, K. Hugl, O. T irkk onen, and V . K oi vunen, “De vice-t o- De vice Communication Underlaying Cellular Communication Systems, International J ournal of Commu- nications, Network and System Sciences , no. 3, pp. 169–178, June 2009. [9] I. Iskandar , I. Setya w an, and H. Nuraini, “Inter -cell interference management technique for multi-ce ll lte-a netw ork, International J ournal of Electrical and Computer Engineering (IJECE) , v ol. 7, no. 5, p. 2696, 2017. [10] T . Peng, Q. Lu, H. W ang, S. Xu, and W . W ang, “Interference A v oidance Mechanisms in the Hybrid Cellular and De vice-to- De vice Systems, in Pr oc. of IEEE 20th International Symposium on P er sonal, Indoor and Mobile Radio Communications , T ok yo, Japan, Sept. 2009, pp. 617–621. [11] M. Meiber gen, “De vice-to-De vice Communicati ons Underlaying a Cel lular Netw ork, Master’ s thesis, Delft Uni v ersity of T echnology , Netherlands, 2011. [12] K. Huang, V . Lau, and Y . Chen, “Spectrum Sharing Between Cellular and Mobile Ad Hoc Netw orks: T ransmission-Capacity T rade -Of f, IEEE J ournal on Selected Ar eas in Communications , v ol. 27, no. 7, pp. 1256–1267, 2009. [13] C. Y u, K. Doppl er , C. Ribeiro, and O. T irkk onen, “Res ou r ce Sharing Optimization for De vice-to-De vice Communication Underlaying Cellular Netw orks, IEEE T r ans. on W ir eless Communications , v ol. 10, no. 8, pp. 2752–2763, 2011. [14] B. Kaufman, J. Lilleber g, and B. Aazhang, “Spectrum Sharing Scheme Between Cellular Users and Ad- Hoc De vice-to-De vice Users, IEEE T r ans. on W ir eless Communications , v ol. 12, no. 3, pp. 1038–1049, March 2013. [15] H. Esmat, M. Elmesala wy , and I. Ibrahim, “Adapti v e Resource Sharing Algorithm for De vice-to-De vice Communications Underlaying Cellular Netw orks, IEEE Communications Letter s , v ol. 20, no. 3, pp. 530– 533, January 2016. [16] X. Lin, J. Andre ws, and A. Ghosh, “Spectrum Sharing for De vice-to-De vice Communication in Cellular Netw orks, IEEE T r ans. on W ir eless Communications , v ol. 12, no. 13, pp. 6727–6740, 2014. [17] C. Y u, O. T irkk onen, K. Doppler , and C. Ribeiro, “Po wer optimization of de vice-to-de vice communication underlaying cellular communication, in Pr oc. 2009 IEEE International Confer ence on Communications (ICC) , Dresden, German y , 2009. [18] H. Min, W . Seo, J. Lee, and D. Hong, “Reliability Impro v ement Using Recei v e Mode Selection in the De vice-to-De vice Uplink Period Underlaying Cellular Netw orks, IEEE T r ans. on W ir eless Communica- tions , v ol. 10, no. 2, pp. 413–418, February 2011. [19] N. Lee, Y . Lin, J. Andre ws, and R. H. Jr ., “Po wer Control for D2D Underlaid Cellular Netw orks: Modeling, Algorithms, and Analysis, IEEE J ournal on Selected Ar eas in Communications , v ol. 33, no. 1, pp. 1–13, No v ember 2014. [20] R. AliHemmati, B. Liang, M. Dong, G. Boudreau, and A. Se yedmehdi, “Multi-channel po wer allocation for de vice-to-de vice communication underlaying cellular netw orks, in Pr oc. 20166 IEEE International Confer ence on Acoustics, Speec h and Signal Pr ocessing (ICASSP) , Shanghai, China, 2016. [21] J. Chen, C. Liu, H. Li, X. Li, H. W ang, and S. Li, “A Cate gorized Resource Sharing Mechanism for De vice-to-De vice Communications in Cellular Netw orks, Mobile Information Systems , v ol. 2016, no. ID 5894752, pp. 1–10, October 2016. [22] Y . Liu and S. Feng, “Interference Pricing for De vice-to-De vice Communications, i n Pr oc. of IEEE ICC 2014 W ir eless Communication Symposium , Sydne y , Australia, June 2014. [23] P . Y uan and G. Bi, “Resource Allocation in D2D-Enabled Cellul ar Netw orks Using Hierarchical Game, in IJECE V ol. 8, No. 5, October 2018: 3767 3777 Evaluation Warning : The document was created with Spire.PDF for Python.