A Dynamic Resource Allocation Scheme for Group Paging in LTE

IEEE INTERNET OF THINGS JOURNAL, VOL. XX, NO. XX

1

A Dynamic Resource Allocation Scheme for Group Paging in LTE-Advanced Networks Ray-Guang Cheng, Senior Member, IEEE, Firas M. Al-Taee, Jenhui Chen, Senior Member, IEEE, and Chia-Hung Wei

Abstract—Group paging is one of the solutions proposed to deal with the radio access network overload problem resulted from bursty machine-type communications (MTC) traffic in LTEAdvanced (LTE-A) networks. In group paging, the base station normally reserves a fixed amount of random access opportunities (RAOs) for the grouped users to perform random access during a paging access interval. However, the number of contending users are quickly decreased and thus, static allocation of RAOs is not efficient. This paper presents a dynamic resource allocation (DRA) scheme which dynamically adjusts the reserved RAOs for group paging based on the estimated number of contending users in each random-access slot. Simulation results demonstrate that, compared with the traditional static allocation scheme, the proposed DRA scheme can improve the utilization of RAOs by 9% under a target access success probability constraint of 90%. Index Terms—dynamic resource allocation, multi-channel slotted ALOHA, overload control, pull-based, random access, resource utilization

I. I NTRODUCTION

M

ACHINE-TYPE communications (MTC) or machineto-machine (M2M) communications is one of the emerging services for 3GPP LTE-Advanced (LTE-A) networks. The deployment of massive MTC devices would generate a huge amount of signaling/data flows which may congest the radio access network (RAN) and the core network [1]. The former issue, RAN overload, is still an open problem due to the lack of an efficient contention resolution mechanism to deal with massive accesses of MTC devices [2]. 3GPP TR 37.868 [3] is the first technical report that elaborates the RAN overload control problem. They classified the RAN overload control schemes into push-based and pull-based approaches [4]. In push-based approaches, MTC devices can automatically transmit data to the network anytime as they wish [5]. In this case, the base station, which is also known as evolved Node B (eNB) in 3GPP, may relieve the congestion by barring some MTC devices [3], [6]–[8], reserving separate This work was supported in part by the Ministry of Science and Technology, Taiwan, R.O.C., under contract MOST 102-2221-E-011-003-MY3. (Corresponding author: Ray-Guang Cheng.) R.-G. Cheng is with the Department of Electronic and Computer Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan. Email: [email protected]. F. M. Al-Taee is with the Circuit Switch Core, Service Delivery Organization, Ericsson Taiwan. Email: [email protected]. J. Chen is with the Department of Computer Science and Information Engineering, School of Electrical and Computer Engineering, College of Engineering, Chang Gung University, Kweishan, Taoyuan 33302, Taiwan. Email: [email protected]. C.-H. Wei is with the Smart Network System Institute, Institute for Information Industry, Taipei, Taiwan. Email: [email protected].

random access channel (RACH) resources for MTC devices [3], [6], [7], [9], dynamically allocating RACH resources [6] (or, random-access opportunities (RAOs) in LTE-A), defining MTC-specific backoff schemes [6], [10], or enforcing MTC devices to perform channel access at different slots [3]. In pull-based approaches, MTC devices can only transmit data to the eNB when they are paged. The performance of push-based approach and the pull-based approach is rarely compared because they are designed for different application scenarios with different design constraints. The push-based approach is designed for traffic belonging to multiple access classes with different arrival processes. In response to congestion, the network operator normally ensures the quality (e.g., access success probability) of high-priority access class by sacrificing the quality of lower-priority access classes. The pull-based approach, however, is designed for traffic belonging to a single access class and with similar and periodic arrival process. In this case, the network operator knows the amount of MTC devices to be activated and thus, can maximize the utilization of the RAOs through dynamically RAO allocation. The RACHs of the orthogonal frequency division multiple access (OFDMA) system are normally modeled as a multichannel slotted ALOHA system [11]. Several push-based mechanisms have been proposed to enhance the performance of slotted ALOHA systems [11]–[17]. In these mechanisms, the authors tried to optimize the performance of the RACHs by adjusting the number of duplicated copies [12], number of allocated channels [13], number of zero backoff retransmissions [14], backoff window size [15], or retransmission probability [16]. Zhou et al. [11] analyzed the throughput and its mean access delay performance of the RACH under both binary exponential and uniform backoff policies in an OFDMA system. Birk and Keren [12] presented a multi-copy transmission policy to increase the successful probability by adjusting the number of duplicated copies. Baron and Birk [13] considered the problem of maximizing the capacity of multichannel slotted ALOHA networks subject to a user specified deadline and a permissible probability of exceeding it. They proposed to maximize the capacity by allocating the number of channels for each group of access attempts according to the offered load. Choi, et al. [14] proposed a fast retrial algorithm with zero backoff to maximize the resource utilization. The authors suggested dynamically adjusting the number of zero backoff retransmissions based on the offered load. Seo and Leung [15] suggested adjusting the backoff window size according to the offered load. Saadawi and Ephremides [16] proposed to adjust the retransmission probability based on

2

the transmission result in the previous slot. Jiang, et al. [17] proposed to adjust the number of reserved channels according to the offered load. It is noted that these mechanisms assumed that the system operates in steady-state and the effectiveness of these proposed mechanisms is highly dependent on the accuracy of the estimated load. Group paging is one of the pull-based RAN overload control schemes proposed by 3GPP to mitigate the RAN overload problem [1]–[4]. In group paging, an eNB assigns a common group identity (GID) to a group of MTC devices and activates the group of MTC devices by sending a single paging message [3], [4]. Upon receiving the paging message, the group of MTC devices will simultaneously transmit a randomly chosen RAOs at the first random-access (RA) slot. A collided MTC device should perform LTE-A random backoff algorithm and retransmit a newly chosen RAO at a new RA slot [18]. This retransmission process repeats until the MTC device successfully transmits its data to the eNB or the retransmission limitation is exceeded [19]. To prevent from interfering the other services, the eNB may reserve dedicated RAOs for the group of devices to perform random access. For simplicity, the dedicated RAOs may be reserved for the whole paging access interval, which is the maximum duration required by the group of UEs to complete their transmission(s). The length of the paging access interval depends on the backoff algorithm and the retransmission limitation [18]. The performance of group paging has been investigated in [20]–[22]. The simulation studies of group paging are given in [20]. Wei, et al. [21] proposed an approximation formula to estimate the number of success and failure users in oneshot random access. Wei, et al. [22] further extended the work in [21] and proposed an analytical model to derive the access success probability, collision probability, cumulative density function (CDF) of access delay, and CDF of preamble transmissions of group paging by considering the implementation constraints of the LTE-A random access procedure. The proposed model takes advantage of a fluid approximation, which permits the system dynamics to be expressed in terms of a difference equation. It was noted in [22] that the number of contending devices in group paging is continuously decreased during the paging access interval and, thus, static allocation of RAOs is not efficient. To address this problem, a dynamic radio resource allocation algorithm for group paging was investigated through computer simulation [23]. This paper extends the work in [23]. We present a dynamic resource allocation (DRA) scheme to allocate RAOs in each RA slot according to the estimated number of contending devices observed in each RA slot. We refine the analytical model presented in [22] and use the model to find the optimal design parameters according to a target quality of service (QoS) constraint [3]. For illustration, the access success probability PS is chosen as the QoS constraint in the rest of this article. The idea of the DRA scheme is to find the optimal value of reserved RAOs in each RA slot to maximize the utilization of RAOs under the target QoS constraint. The idea itself is not new. However, the effectiveness of the DRA scheme relies on precise knowledge of the number of contending devices in each RA slot, which is hard to be predicted.


This paper presents a new formula to estimate the number of contending devices in each RA slot and uses the refined model to optimize the resource allocation. Considering the implementation limitation of LTE-A networks, we further propose two approaches to notify the dynamically allocated RAOs in each RA slot based on the capability of the MTC devices being paged. More specifically, the major contributions of this paper are highlighted as follows. 1) We propose a DRA scheme to dynamically allocate RAOs in each RA slot for group paging. An analytical model is then presented to determine the optimal number of RAOs in each RA slot under a target access success probability constraint. 2) We propose two approaches to announce the optimal number of RAOs reserved for a group of UEs. The two approaches can be chosen as a trade-off between the UE’s computational power and the utilization of RAOs. The rest of this paper is organized as follows. The system model is given in Section II. The refined analytical model for group paging is then elaborated. Section III presents the proposed DRA scheme. Simulation results are presented in Section IV to verify the effectiveness of the proposed scheme. Section V draws the conclusion and future work. II. S YSTEM M ODEL This paper considers an MTC server which utilizes the LTEA network to communicate with a group of M MTC devices in a cell. The M MTC devices share a common GID such that an eNB can send a single group paging message to notify the M MTC devices to communicate with the MTC server whenever needed. In LTE-A, the random-access resource is determined in terms of RAOs. The total number of RAOs provided by an eNB in a RA slot is equal to the number of frequency bands in each RA slot multiplied by the number of RA preambles [18]. We consider the case that all M MTC devices synchronously transmit their access attempts at the upcoming RA slot immediately after receiving the group paging message. It is assumed that the eNB reserves Ri RAOs at the ith RA slot (1 ≤ Ri ≤ N, 1 ≤ i ≤ Imax ), where N is the maximal number of RAOs in an RA slot and Imax is the total number of RA slots in a paging access interval [22]. The value of Imax can be easily obtained by [22] TRAR + WRAR + WBO , (1) Imax = 1 + (NP Tmax − 1) TRA REP where NP Tmax is the maximal number of preamble transmission; TRAR is the processing time required by the eNB to detect the transmitted RA requests (unit: subframe); WRAR is the length of the RA response window (unit: subframe); WBO is the backoff window size (unit: subframe); and TRA REP (unit: subframe) is the time interval between two RA slots. The utilization of RAOs, denoted as U , is the ratio of the average number of successfully accessed RAOs to the total number of reserved RAOs. U is given by [22] PImax PNP Tmax Mi,S [n] U = i=1 Pn=1 , (2) Imax i=1 Ri

CHENG, AL-TAEE, CHEN, WEI

3

where Mi,S [n] is the average number of MTC devices which transmit their nth preamble at the ith RA slot and successfully complete the RA procedure. In this paper, the performance metric of access success probability, denoted as PS , defined by 3GPP [3] is chosen as an example of the target QoS constraint. PS is the percentage of devices which successfully complete the RA procedure, which is given by [22, (14)] PImax PNP Tmax Mi,S [n] n=1 . (3) PS = i=1 M To obtain Mi,S [n], it can be obtained by replacing R by Ri in [22, (4)] and is given by  M − i   Mi [n]e Ri pn ,    NP Tmax   X M  − i   Mi [n]e Ri pn ≤ NU L , (4a) if     n=1 M Mi,S [n] = − i Mi [n]e Ri pn   NU L ,   NP Tmax  X  Mi  −   Mi [n]e Ri pn      n=1 otherwise. (4b) and is equivalent to M

Mi,S [n] ≡

Mi [n]e

− Ri

NP Tmax

X

i

pn

M − Ri i

Mi [n]e

MACK,i ,

(5)

pn

n=1

where Mi [n] is the number of MTC devices transmitting their nth preamble at the ith RA slot; Mi is the average number of contending MTC devices at the ith RA slot; NU L is the maximal number of MTC devices that can be acknowledged within the RA response window; pn is the detection probability of the nth preamble transmission; and MACK,i is the total number of acknowledged MTC devices at the ith RA slot and is equal to  NP Tmax  X M  − i   Mi [n]e Ri pn ,     n=1 NP Tmax MACK,i ≡ (6) X M − i   if Mi [n]e Ri pn ≤ NU L ,    n=1   N , otherwise. UL Mi is the summation of Mi [n] over n and is given by [22, (3)] NP Tmax X Mi = Mi [n]. (7) n=1

Mi [n] can be obtained by replacing R by Ri in [22, (6)] and is given by Mi [n] =

K max X

αk,i Mk,F [n − 1]

k=Kmin

=

K max X k=Kmin

αk,i (Mk [n − 1] − Mk,S [n − 1]),

(8)

where αk,i is the percentage of MTC devices which fail at the kth RA slot and retransmit at the ith RA slot; Mk,F [n − 1] is the number of devices which fail their (n − 1)th transmission at the kth RA slot; and Kmin and Kmax are the first and last RA slots from which an MTC device can retransmit its ranging code at the ith RA slot, respectively. The number of failed MTC devices is equal to the number of contending MTC devices minus the number of successful MTC devices (i.e., Mk,F [n] = Mk [n] − Mk,S [n]); Kmin , Kmax , and αk,i can be obtained from [22, (8)–(10)], respectively. The effectiveness of the DRA scheme is highly dependent on the estimated value of Mi . However, it is found that the estimation error of MACK,i in (6) becomes significant when MACK,i approaches NU L . To correct the error, we replace ˜ ACK,i , which is given MACK,i in (6) by a corrected term M by N Ri UL X X ˜ ACK,i = NU L Pi (k), (9) kPi (k) + M k=1

k=NU L

where Pi (k) is the probability density function that the eNB detects k MTC devices at the ith RA slot. Note that the eNB can only detect up to Ri MTC devices at the ith RA slot. Therefore, the maximal value of k is Ri . It is found that Pi (k) can be approximated by a Poisson distribution with mean λi , where λi is the average number of detected MTC devices at the ith RA slot. It is equal to the sum of devices transmitting their nth ranging code at the ith RA slot and can be detected by the eNB. From [21, (6)], we can have NP Tmax

λi =

X

M

− Ri

Mi [n]e

i

pn .

(10)

n=1

Substituting Pi (k) by a Poisson distribution with mean λi in (9), we can have ˜ ACK,i = M

N UL X k=1

k

Ri X (λi )k e−λi (λi )k e−λi NU L + . (11) k! k! k=NU L

III. DYNAMIC R ESOURCE A LLOCATION S CHEME A. DRA Scheme In this section, the DRA scheme is introduced to find the optimal value of Ri to maximize the utilization of RAOs under a target QoS constraint. A common approach for finding an optimal Ri for the ith RA slot is to perform a global search over all possible values of Ri for all Imax RA slots. However, the dimension of the search space is N Imax , which cannot be found for large N and Imax . The eNB cannot use a single group paging message the optimal values of Ri for the whole paging access interval. Therefore, we proposed a heuristic approach to find the proper values of Ri . Leveraging the knowledge of slotted ALOHA, we suggest setting Ri proportional to the estimated number of contending devices (Mi ). From (6), it is found that the total number of acknowledged devices at the ith RA slot, MACK,i , is bounded by NU L . Therefore, it is not efficient to set a very high value of Ri since the extra UEs are detected but not acknowledged. Since Ri must be an integer with an upper bound N , we have Ri = dmin(N, µMi , Ri,max )e ,

(12)

4


1

60 Optimal Ri Approximated Ri

50

0.9 0.8 0.7

Ri

CDF of Pi(k)

40

30

20

0.6 0.5 0.4 0.3

Original (λ26)

0.2

Approximated (λ26)

0.1

Original (λ18)

10

0

0 0

5

10

15

20 25 30 35 the i-th RACH

40

45

50

55

where µ is the smallest constant chosen to attain the target QoS constraint and Ri,max is the maximal value of Ri such that the average number of detected devices at the ith RA slot is equal to NU L . Therefore, Ri,max is determined by setting λi = NU L in (10) and is given by   Ri,max

  Mi . ! PNP Tmax   M [n]p i n n=1   NU L

0

5

10

15

20

25

30

35

40

k Fig. 2. CDF of Poisson function and Pi (k) for λ18 and λ26 .

ˆ i when M = 400 and PQoS = 0.95. Fig. 1. Optimal Ri and R

  =    ln 

Approximated (λ18)

(13)

Note that (11) only applies for Mi > NU L since Ri = ∞ if we want to ensure λi = NU L for Mi ≤ NU L . Consider the case that the network utilizes the access success probability PS as the QoS constraint. Let PQoS be the target QoS constraint (i.e., the low bound of PS ). The solution of µ in (12) satisfying PS ≥ PQoS can then be numerically found. B. Implementation Consideration In the implementation, two approaches are adopted. The first approach is suitable for the group of UEs with high computation power. In this case, the eNB will carry µ, NP Tmax , NU L , M , and Imax in the group paging message. The group of UEs can use these parameters to calculate Mi through (1) to (7) and use Mi to derive Ri from (12) in real-time. The second approach is suitable for the group of UEs with low computation power. In this approach, the eNB has to find a simple function which approximates the optimal value of Ri in (12) and uses the group paging message to inform the group of UE the parameters of the simple function. An approach to find the simple function is illustrated as below. Fig. 1 shows an example illustrating the optimal value of Ri and the approximated Ri for M = 400 and PQoS = 0.95. From (12), it is found that R1 = N = 54 [3] and R2 = 0 for all M > N . We want to use a constant and an exponential î function to approximate Ri . The approximation function R

can be characterized by four parameters of a, b, c, and d. That is,  b X R i  ≡ a, for 3 ≤ i < b, î = b−3 R (14) i=3   d·i c·e , for b ≤ i ≤ Imax . The parameter b in (14) can be obtained by finding the integer î − value of i which minimizes the approximation error of |R Ri |. Parameters c and d are obtained through curve fitting of the function of Ri for b ≤ i < Imax . For example, the ˆ i in Fig. 1 are a = 47, b = 22, c = 595.3, parameters of R and d = −0.1182. IV. S IMULATION R ESULTS Monte-Carlo-based computer simulations are conducted to verify the accuracy of the analytical model and the effectiveness of the DRA scheme. The RACH parameters defined in [3, Table 6.2.2.1.1] are used as a baseline throughout all simulation experiments. All simulation results shown in the following figures are obtained by averaging 106 samples. Each sample represents the result obtained by observing a group paging of M devices during a full paging access interval containing Imax RA slots. In all figures, we use lines and symbols to represent analytical and simulation results, respectively. In the simulation, two scenarios are discussed. Scenario I is set to ˜ ACK,i given verify the effectiveness of the modified term M in (11). Scenario II is designed to verify the accuracy of the proposed analytical model and the effectiveness of the DRA scheme. In Scenario I, the effectiveness of the Poisson approximation for Pi (k) is first demonstrated. The cumulative distribution function (CDF) of Pi (k) for M = 400, NU L = 15, and µ = 1 is shown in Fig. 2. In this figure, we use lines and symbols to represent ‘CDF of Poisson function’ and ‘CDF of Pi (k),’ respectively. The values of λi of the 18th and 26th RA slots are obtained from (10) and are λ18 = 16 and λ26 = 12, respectively. As shown in Fig. 2, the CDF


5

TABLE I E STIMATION E RROR OF ACCESS S UCCESS P ROBABILITY PS

# acknowledged MTC devices at the i-th RA slot

M Analysis µ = 0.6 Simulation Error Analysis µ = 1.0 Simulation Error Analysis µ = 1.8 Simulation Error

100 0.911839 0.883850 0.031667 0.996807 0.968880 0.028824 0.999998 0.996340 0.003671

200 0.873141 0.865270 0.009097 0.988891 0.978650 0.010464 0.999836 0.997130 0.002714

300 0.862948 0.870180 0.008311 0.976922 0.978703 0.001820 0.997345 0.996220 0.001129

400 0.838436 0.856400 0.020976 0.951802 0.963233 0.011867 0.984658 0.988657 0.004045

500 0.794558 0.812784 0.022424 0.894724 0.908432 0.015090 0.928263 0.939142 0.011584

600 0.713243 0.724792 0.015934 0.770399 0.782858 0.015915 0.793609 0.805865 0.015209

700 0.562072 0.568036 0.010499 0.595735 0.603017 0.012076 0.612625 0.619899 0.011734

800 0.392590 0.395549 0.007481 0.417395 0.419482 0.004975 0.430660 0.433289 0.006068

900 0.270470 0.271270 0.002949 0.290425 0.291578 0.003954 0.300861 0.301287 0.001414

1000 0.191223 0.191650 0.002228 0.208842 0.209660 0.003902 0.218261 0.219300 0.004738

1

16 Simulation Analysis (Modified) Analysis (Old)

14

0.9 0.8

12

0.7

10

0.6 PS

8

0.5 0.4

6

Simulation (µ = 0.6) Simulation (µ = 1.0) Simulation (µ = 1.8) Analysis (µ = 0.6) Analysis (µ = 1.0) Analysis (µ = 1.8)

0.3

4

0.2

2

0.1 0

0 0

5

10

15

20 25 30 35 the i-th RACH

40

45

50

100

55

Fig. 3. Total number of acknowledged MTC devices at the ith RA slot for M = 400.

200

300

400

500

600

700

800

900 1,000

M

Fig. 4. Comparison of access success probability PS for various values of µ. 0.35

of Poisson distribution well approximates the CDF of Pi (k) as expected. Fig. 3 shows the results for the total number of acknowledged MTC devices at the ith RA slot for M = 400. In this figure, ‘old model’ refers to the analytical results of MACK,i obtained from [22, (4)]; ‘modified model’ represents ˜ ACK,i obtained from (11); and the the analytical results of M simulation results are obtained by counting the acknowledged devices over the paging access interval during group paging. It shows that the modified model is much more accurate than the old model in approximating the simulation results of acknowledged MTC devices. Scenario II is designed to verify the accuracy of the proposed analytical model and the effectiveness of the DRA scheme. The accuracy of the proposed analytical model is first demonstrated through comparing analytical and simulation results for M = 100, 200, . . . , 1000. Fig. 4 shows the analytical and simulation results of the access success probability PS for various values of µ. The estimation error of the analytical model is shown in Table I. These data validate the correctness of the analytical model. Fig. 5 shows the average utilization of RAOs, U , for various values of µ. It can be found in Figures 4 and 5 that the accuracy of the analytical model decreases if a smaller µ is chosen. A smaller values of µ implies less

0.3 0.25 0.2 U 0.15

Simulation (µ = 0.6) Simulation (µ = 1.0) Simulation (µ = 1.8) Analysis (µ = 0.6) Analysis (µ = 1.0) Analysis (µ = 1.8)

0.1 0.05 0 100

200

300

400

500

600

700

800

900 1,000

M

Fig. 5. Comparison of average utilization of RAOs for various values of µ.

RAOs are reserved, which results in in a higher approximation error in the Poisson approximation. However, as illustrated in Table I, the maximal estimation error of PS is less than 3.3%, which demonstrates the accuracy of the new model. In the following, the efficiency of the DRA scheme for two

6


1.5

TABLE III PARAMETERS R EQUIRED TO A PPROXIMATED Ri FOR M FROM 100 TO 450 AND PQoS = 0.95

PQoS = 0.95 PQoS = 0.9

1.4

Optimal value of µ

1.3

M 100 150 200 250 300 350 400 450

1.2 1.1 1 0.9

a 14 25 35 40 44 46 47 47

b 5 6 6 9 12 16 22 29

c 23.81 42.03 61.03 84.8 121.9 215.3 595.3 755.5

d −0.069 −0.07619 −0.08016 −0.08235 −0.08644 −0.09669 −0.1182 −0.1773

0.8 0.7

0.32

0.6

0.3

0.5 50

100

150

200

250

300

350

400

450

0.28

500

M

0.26

Fig. 6. Optimal value of µ for PQoS = 0.9 and PQoS = 0.95. U

TABLE II PARAMETERS R EQUIRED TO A PPROXIMATED Ri FOR M FROM 100 TO 450 AND PQoS = 0.90

0.24

Static DRA Old Dynamic Approximated Ri

0.22 0.2

M 100 150 200 250 300 350 400 450

a 12 18 25 32 42 47 48 49

b 5 4 4 4 9 12 17 22

c 19.72 32.17 45.33 58.15 83.44 114.6 197.8 448.1

d −0.05783 −0.06019 −0.06248 −0.06323 −0.06979 −0.07397 −0.08482 −0.101

target QoS constraints of PS = 0.9 and PS = 0.95 are illustrated based on the proposed analytical model. Fig. 6 shows the optimal value of µ for PQoS = 0.9 and PQoS = 0.95 determined by the DRA scheme. The optimal value of µ for M = 500 and PQoS = 95% is not shown on the figure for data presentation clarity. Table II and Table III show the parameters required to approximate resource allocation for PQoS = 0.9 and 0.95 respectively. All parameters are found through the approximation algorithm presented in Section III-B. Figures 7 and 8 demonstrate the utilization of RAOs, U , of group paging. In the two figures, ‘static,’ ‘old dynamic,’ ‘DRA,’ and ‘Approximated Ri ,’ refer to the static resource allocation scheme, the dynamic resource allocation scheme proposed in [23], the DRA scheme proposed in Section III, and the proposed DRA scheme with approximated Ri (i.e., ˆ i given in Section III-B), respectively. To ensure a fair comR parison, the minimal value of Ri required to ensure the target QoS in chosen in both ‘static’ and ‘old dynamic’ schemes. Fig. 7 shows that the DRA scheme achieves better performance that the one presented in [23] and a much better performance than that of the static resource allocation. Compared with the DRA scheme, the utilization of RAOs in DRA scheme with approximated Ri scheme is slightly degraded. However, it requires lower signaling overhead and computation complexity and thus, is a preferred choice for implementation.

0.18 0.16 100

150

200

250

300

350

400

450

M

Fig. 7. Average utilization of RAOs for group paging with different resource allocation algorithms (PQoS = 0.90).

V. C ONCLUSION In this paper, we proposed a DRA scheme to optimize resource utilization for group paging in MTC networks. The proposed DRA scheme utilizes the analytical model presented in [22] to estimate the behavior of MTC devices in group paging. However, it is found that existing model has noticeable error under DRA. Therefore, we presented a new formula to eliminate estimation error and used the refined model to optimize the resource allocation. The accuracy of the refined model was verified via computer simulation. However this paper has modified that analytical model to support dynamic resource allocation. Simulation results showed that resource utilization is improved by at least 65% of that in the static group paging described in [22]. We further proposed a fourparameter function to approximate the DRA results to reduce the computation and signaling complexity in the implementation. Simulation results showed the four-parameter function can greatly reduce the signaling overhead at a cost of slightly degraded utilization. Future research may consider replacing the constant relativity between the number of contending devices and the allocated resources by a function of time. Also, other methods can be proposed to encounter the implementation limitations with less effect over the optimized resource utilization. One


7

0.32 0.3 0.28 0.26 U

0.24

Static DRA Old Dynamic Approximated Ri

0.22 0.2 0.18 0.16 100

150

200

250

300

350

400

450

M

Fig. 8. Average utilization of RAOs for group paging with different resource allocation algorithms (PQoS = 0.95).

more related topic that is worthy to be considered is overlapped group paging, where a group is paged during the paging access interval of another group, making use of the light load at the end of the paging access interval. R EFERENCES [1] S. Y. Lien, K. C. Chen, and Y. Lin, “Toward Ubiquitous Massive Accesses in 3GPP Machine-to-Machine Communications,” IEEE Commun. Mag., vol. 50, no. 4, pp. 66–74, Apr. 2011. [2] M. Y. Cheng, G. Y. Lin, H. Y. Wei, and C. C. Hsu, “Overload Control for Machine-Type-Communications in LTE-Advanced System,” IEEE Commun. Mag., vol. 50, no. 6, pp. 38–45, June 2012. [3] 3GPP TR 37.868, “Technical Specification Group Radio Access Network; Study on RAN Improvements for Machine-Type Communications,” v.11.0.0, Sep. 2011. [4] 3GPP R2-104870, “Pull Based RAN Overload Control,” Huawei and China Unicom, RAN2 #71, Aug. 2010. [5] A. Laya, L. Alonso, and J. Alonso-Zarate, “Is the Random Access Channel of LTE and LTE-A Suitable for M2M Communications? A Survey of Alternatives,” IEEE Commun. Surveys & Tuts., vol. 16, no. 1, pp. 4–16, 1st quarter 2014. [6] 3GPP TSG R2-104662, “MTC Simulation Results with Specific Solutions,” ZTE, RAN WG2 #71, Madrid, Spain, Aug. 2010. [7] J.-P. Cheng, C.-H. Lee, and T.-M. Lin, “Prioritized Random Access with Dynamic Access Barring for RAN Overload in 3GPP LTE-A Networks,” in Proc. IEEE GLOBECOM 2011, pp. 368–372, Houston, TX, Dec. 2011. [8] S.-Y. Lien, T.-H. Liau, C.-Y. Kao, and K.-C. Chen, “Cooperative Access Class Barring for Machine-to-Machine Communications,” IEEE Trans. Wireless Commun., vol. 11, no. 1, pp. 27–32, Jan. 2012. [9] K.-D. Lee, S. Kim, and B. Yi, “Throughput Comparison of Random Access Methods for M2M Service over LTE Networks,” in Proc. IEEE GLOBECOM 2011, pp. 373–377, Houston, TX, Dec. 2011. [10] X. Yang, A. Fapojuwo, and E. Egbogah, “Performance Analysis and Parameter Optimization of Random Access Backoff Algorithm in LTE,” in Proc. IEEE VTC2012-Fall, pp. 1–5, Québec, Canada, Sep. 2012 [11] P. Zhou, H. Hu, H. Wang, and H. H. Chen, “An Efficient Random Access Scheme for OFDMA Systems with Implicit Message Transmission,” IEEE Trans. Wireless Commun., vol. 7, no. 7, pp. 2790–2797, Jul. 2008. [12] Y. Birk and Y. Keren, “Judicious Use of Redundant Transmissions in Multichannel ALOHA Networks with Deadlines,” IEEE J. Sel. Areas Commun., vol. 17, no. 2, pp. 257–269, Feb. 1999. [13] D. Baron and Y. Birk, “Multiple Working Points in Multichannel ALOHA,” in Proc. Wireless Networks, vol. 8, pp. 5–11, Feb. 2002.

[14] Y. J. Choi, S. Park, and S. Bahk, “Multichannel Random Access in OFDMA Wireless Network,” IEEE J. Sel. Areas Commun., vol. 24, no. 3, pp. 603–613, Mar. 2006. [15] J. B. Seo and V. C. Leung, “Design and Analysis of Backoff Algorithms for Random Access Channels in UMTS-LTE and IEEE 802.16 Systems,” IEEE Trans. Veh. Technol., vol. 60, no. 8, pp. 3975–3989, Oct. 2011. [16] T. N. Saadawi and A. Ephremides, “Analysis, Stability, and Optimization of Slotted ALOHA with a Finite Number of Buffered Users,” IEEE Trans. Automatic Control, vol. 26, no. 3, pp. 680–689, Jun. 1981. [17] F. Jiang, H. Tian, and P. Zhang, “An Adaptive Random Access Strategy for Multi-Channel Relaying Networks,” Science in China Series F: Information Sciences, vol. 52, no. 12, pp. 2406–2414, Dec. 2009. [18] 3GPP TS 36.321, “Evolved Universal Terrestrial Radio Access (EUTRA) Medium Access Control (MAC) Protocol Specification,” v.9.3.0, Jun. 2010. [19] R1-060584, “E-UTRA Random Access,” Ericsson. [20] 3GPP R2-113198, “Further Analysis of Group Paging for MTC,” ITRI, RAN2#74, May 2011. [21] C.-H. Wei, R.-G. Cheng, and S.-L. Tsao, “Modeling and Estimation of One-Shot Random Access for Finite-User Multichannel Slotted ALOHA Systems,” IEEE Communications Letters, vol. 16, no. 8, pp. 1196–1199, Aug. 2012. [22] C.-H. Wei, R.-G. Cheng, and S.-L. Tsao, “Performance Analysis of Group Paging for Machine-Type Communications in LTE Networks,” IEEE Trans. Veh. Tech., vol. 66, no. 7, pp. 3371–3382, Sep. 2013. [23] C.-H. Wei, R.-G. Cheng, and F. M. Al-Taee, “Dynamic Radio Resource Allocation for Group Paging Supporting Smart Meter Communications,” in Proc. IEEE Third Int. Conf. Smart Grid Commun., pp. 659–663, Nov. 2012.

Ray-Guang Cheng (S’94–M’97–SM’07) received the B.E., M.E., and Ph.D. degrees in communication engineering from National Chiao Tung University, Taiwan, in 1991, 1993, and 1996, respectively. From 1997 to 2000, he was with Advance Technology Center, Computer and Communication Laboratories, Industrial Technology Research Institute (ITRI), Taiwan, as a researcher and a project leader. From 2000 to 2003, he joined BenQ Mobile System Inc., Taiwan, as a senior manager of R&D division. He is currently a Professor with the Department of Electronic and Computer Engineering, National Taiwan University of Science and Technology (NTUST), Taiwan. His research interests include multi-hop wireless networks and machine-to-machine communications. Dr. Cheng is a senior member of IEEE and Phi Tau Phi scholastic honor society. He holds IEEE Wireless Communication Professional (WCP) certification and 18 US patents; and has published more than 90 international journal and conference papers and more than 30 IEEE/3GPP standard contributions. He led the 3G Protocol project and his team was named Top Research Team of the Year by ITRI in 2000. He received the Best Industrial-based Paper Award from Ministry of Education in 1998; Advanced Technologies Award from Ministry of Economic Affairs in 2000; and Teaching Award, Research Award, and Excellence in Counseling Award from NTUST in 2006, 2009, and 2011, respectively.

Firas M. Al-Taee earned his master degree in electronic and computer engineering from National Taiwan University of Science and Technology (NTUST), Taipei, Taiwan in 2013. He then worked as an RD engineer for Accton Technology Group until 2015. Since early 2015, he has been working as an IMS Core Network engineer in Ericsson, Taiwan until current day.

8


Jenhui Chen (S’99–M’03–SM’14) received the B.S. and Ph.D. degrees in Computer Science and Information Engineering (CSIE) from Tamkang University, Taipei, Taiwan, R.O.C. in July 1998 and January 2003, respectively. Since 2003, he has been with the Department of CSIE, School of Electrical and Computer Engineering, College of Engineering, Chang Gung University, where he is currently a professor and a team leader of the High Speed Intelligent Center (HSIC), Chang Gung University. His main research interests include design, analysis, and implementation of computer and communication protocols, wireless networks, software-defined networks (SDN), cloud computing, big data, augmented reality (AR), and interconnects network-on-chip (NoC) for multi-core and multi-processor system-on-chip (MPSoC) design. Dr. Chen is a senior member of IEEE. He is currently an editor of The Scientific World Journal and Advances in Computer Engineering and an associate editor of Cogent Engineering. Dr. Chen served as the technical program committee (TPC) member of IEEE AINA 2011, IEEE ICCCN 2014, 2015, and IEEE ICC 2015. He also served as a reviewer for many famous academic journals which are organized by ACM, Elsevier, IEEE, and Springer.

Chia-Hung Wei earned his Ph.D. degree in electronic and computer engineering from National Taiwan University of Science and Technology (NTUST), Taipei, Taiwan in 2013. Dr. Wei is currently a senior engineer in Smart Network System Institute (SNSI), Institute for Information Industry (III), Taipei, Taiwan, R.O.C. His research interests include machine-type communications, multichannel slotted ALOHA, and performance analysis of cellular networks.