versatile access control for massive iot: throughput, latency,...

Versatile Access Control for Massive IoT:Throughput, Latency, and Energy Efficiency

Han Seung Jang ,Member, IEEE, Hu Jin , Senior Member, IEEE,

Bang Chul Jung , Senior Member, IEEE, and Tony Q. S. Quek , Fellow, IEEE

Abstract—In this paper, we propose a novel access control (AC) mechanism for cellular internet of things (IoT) networks with massive

devices, which effectively satisfies various performance metrics such as access throughput, access delay, and energy efficiency. Basic

idea of the proposed AC mechanism is to adjust access class barring (ACB) factor according to performance targets. For a given

performance target, we derive the optimal ACB factors by considering not only the conventional preamble collision detection technique

but also the early preamble collision detection technique, respectively. In addition, the proposed AC mechanism considers overall radio

resources to optimize the ACB factor, which includes the number of preambles, random access response (RAR) messages, and

physical-layer uplink shared channels (PUSCHs), while most conventional ACB schemes consider only the number of preambles. In

particular, the proposed AC mechanism is illustrated with two representative performance metrics: latency and energy efficiency.

Through extensive computer simulations, it is shown that the proposed versatile AC mechanism outperforms the conventional ACB

schemes in terms of various performance metrics under diverse resource constraints.

Index Terms—Internet of things, low-latency, energy-efficiency, access class barring (ACB), ACB factor, backlog estimation

Ç

1 INTRODUCTION

1.1 Background

TYPICAL communication networks, which are opti-mized for human-to-human (H2H) communications,

are evolving to support machine-to-machine (M2M) com-munications or internet of things (IoT) services [1]. Inparticular, 5G cellular wireless networks have beendeveloped and standardized under careful considerationon IoT services in the third generation partnership proj-ect (3GPP) and international telecommunication union(ITU) [2], [3]. For supporting massive connections, theIoT networks may require various technologies includingedge computing [4], artificial intelligence (AI) [5], etc.Optimizing M2M communication protocols is anotherinteresting topic to satisfy various quality of service(QoS) requirements in the IoT networks [6], [7]. Forexample, in smart city, autonomous vehicles requirevery low latency communications with nearby vehicles,

while temperature sensors require an ultra-high energyefficiency (EE) to prolong their battery life since the bat-tery lifetime is required to be around 10 years. In partic-ular, Sivanathan et al. [8] proposed a robust frameworkfor IoT device classification using network traffic charac-teristics based on machine learning schemes.

Existing commercial cellular networks may encountercritical bottlenecks if they are adopted for massive IoTnetworks without significant modifications. First of all,the amount of radio resources in the cellular network isvery limited to accommodate a rapidly growing numberof IoT devices, and thus careful resource management isrequired [9], [10], [11]. Andrade et al. [9], [10] addressedthat the control channels including physical downlinkchannel (PDCCH) may be limited to support overloadcontrol, and they proposed a prioritized control resourcescheduling algorithm considering both of H2H and M2Mcommunications. Xia et al. [11] provided a comprehen-sive technical survey on the resource management tech-niques for M2M communications in commercial LTE/LTE-A cellular networks.

1.2 Related Works

Technical challenges may come from massive nodes that areactivated simultaneously due to bursty traffic. In general,the bursty traffic pattern causes severe congestion or over-load problem in a radio access network (RAN). To over-come the congestion problem, several overload controlschemes have been proposed [12], [13], [14], [15], [16], [17],[18], [19]. Among them, the access class barring (ACB)scheme has been considered as a promising technologysince it effectively controls bursty traffic in cellular M2Mnetworks [14]. In particular, the dynamic access class

� H. S. Jang is with the School of Electrical, Electronic Communication, andComputer Engineering, Chonnam National University, Yeousu 59626,Republic of Korea. E-mail: [email protected].

� H. Jin is with the Division of Electrical Engineering, Hanyang University,Ansan 15588, Republic of Korea. E-mail: [email protected].

� B. C. Jung is with the Department of Electronics Engineering, ChungnamNational University, Daejeon 34134, Republic of Korea.E-mail: [email protected].

� T. Q. S. Quek is with the Singapore University of Technology and Design,Singapore 487372, and also with the Department of Electronic Engineering,Kyung Hee University, Yongin 17104, South Korea.E-mail: [email protected].

Manuscript received 1 Sept. 2018; revised 23 Mar. 2019; accepted 26 Apr.2019. Date of publication 2 May 2019; date of current version 1 July 2020.(Corresponding authors: Hu Jin and Bang Chul Jung.)Digital Object Identifier no. 10.1109/TMC.2019.2914381

1984 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 19, NO. 8, AUGUST 2020

1536-1233� 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See ht _tps://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Chungnam National University. Downloaded on July 12,2020 at 07:06:12 UTC from IEEE Xplore. Restrictions apply.

https://orcid.org/0000-0002-9024-8952

https://orcid.org/0000-0002-9024-8952

https://orcid.org/0000-0002-9024-8952

https://orcid.org/0000-0002-9024-8952

https://orcid.org/0000-0002-9024-8952

https://orcid.org/0000-0002-3505-6843

https://orcid.org/0000-0002-3505-6843

https://orcid.org/0000-0002-3505-6843

https://orcid.org/0000-0002-3505-6843

https://orcid.org/0000-0002-3505-6843

https://orcid.org/0000-0002-4485-9592

https://orcid.org/0000-0002-4485-9592

https://orcid.org/0000-0002-4485-9592

https://orcid.org/0000-0002-4485-9592

https://orcid.org/0000-0002-4485-9592

https://orcid.org/0000-0002-4037-3149

https://orcid.org/0000-0002-4037-3149

https://orcid.org/0000-0002-4037-3149

https://orcid.org/0000-0002-4037-3149

https://orcid.org/0000-0002-4037-3149

mailto:

mailto:

mailto:

mailto:

barring (D-ACB) schemes have been actively studied inorder to support a varying number of access nodes in IoTnetworks [17]. The D-ACB schemes estimate the number ofbacklogged nodes to attempt access at each time slot, andthen it generates an optimal ACB factor, also known asaccess probability, to allow nodes to send a preamble viaphysical random access channel (PRACH), by consideringthe available resources. Jin et al. [18] proposed a recursivepseudo-Bayesian ACB scheme in order to lower total servicetime of a bursty traffic, and a practical backlogged estima-tion scheme based on the number of undetected preamblesat the first step of the random access procedure. On theother hand, Tavana et al. [15] proposed a backlog estimationmethod based on Kalman filtering for the adaptive ACBscheme, where it is shown that total service time becomesclose to that with the optimal ACB scheme that exactlyknows the number of backlogged nodes. Duan et al. pro-posed a dynamic preamble allocation scheme and an itera-tive algorithm to adaptively update the ACB factor forM2M devices. Vural et al. [19] proposed an adaptive pream-ble slicing algorithm, called multi-preamble random access,to dynamically determine the preamble set size accordingto different load conditions and service classes. However,most of existing studies focused on the dynamic access con-trol mechanisms to meet a single performance target suchas maximization of the number of successful access orequivalently minimization of access delay and total servicetime. In addition, they only considered the preamble resour-ces at the first step of RA procedure even though otherradio resources are involved in the overall RA procedure.

EE is one of the most important performance metricsespecially for battery-powered IoT devices, and thus manytechniques have been proposed in literature [20], [21], [22],[23], [24], [25], [26], [27], [28], [29]. Ho et al. [20] proposed ajoint massive access control and resource allocation schemeto improve the EE, which adopts machine node grouping,coordinator selection, and optimal power allocation forcoordinators. Zhang et al. [26] proposed an integratedenergy efficient system for 5G IoT networks, which holisti-cally combines the wireless and wired parts together inorder to optimize the EE of the system. Azari and Miao [28]proposed a resource allocation framework based on themax-min lifetime-fairness, which exploits the informationon the battery time of devices. Yang et al. [27] proposed anenergy efficient resource allocation scheme with energy har-vesting technology in time division multiple access andnon-orthogonal multiple access strategies. However, theeffect of the access control technique on the EE have beenrarely investigated in massive IoT networks to the best ourknowledge.

1.3 Contributions

In this paper, we propose a versatile access control mecha-nism that effectively satisfies various performance require-ments including access throughput, access delay, and theEE for massive IoT networks. To obtain the ACB factor forsatisfying latency or EE requirement, the proposed mecha-nism considers all the radio resources involved in the over-all RA procedure: preambles on PRACH, RAR messages onphysical downlink shared channel (PDSCH) or PDCCH,and uplink data resources on physical uplink shared

channel (PUSCH). In addition to the conventional preamblecollision detection (C-PCD) technique, we also exploit theearly preamble collision detection (E-PCD) technique at thefirst RA procedure [30], [31]. Furthermore, we propose acomprehensive access control mechanism which can simul-taneously support two groups of devices requesting low-latency and ultra-low power consumption services, respec-tively. Simulation results show that the proposed versatileAC mechanism efficiently satisfies various performancerequirements under various resource constraints and it out-performs the existing AC schemes. The main contributionsof this paper can be summarized as follows:

� We propose a versatile AC mechanism to efficientlycontrol random access traffic from a massive numberof IoT nodes while achieving various performancerequirements such as low-latency and energy-efficiency.

� We derive the ACB factors for low-latency accessand energy-efficient access with the conventionalpreamble collision detection technique and the earlypreamble collision detection technique, respectively.

� We also propose a comprehensive AC mechanismin order to simultaneously support both of low-latency access group and energy-efficient accessgroup.

The rest of this paper is organized as follows. In Section 2,we describe a system model considered in this paper. InSection 3, we derive preliminary probabilities of collision-free and collided preambles which are utilized to optimizethe ACB factor of the proposed mechanism. In Section 4, wepropose two types of AC mechanisms: low-latency AC andenergy-efficient AC mechanisms. Then, we explain the esti-mation algorithm for the number of backlogged nodes inSection 5. The proposed AC mechanisms are validated viaextensive computer simulations in Section 6. Finally, con-clusions are drawn in Section 7.

2 SYSTEM MODEL

Fig. 1 shows the system model of a single cell massive IoTnetwork which is considered in this paper. We assume thatthe eNodeB controls uplink access from two representativegroups1, each of which requires a certain performance

Fig. 1. System model of a massive IoT network with versatile accesscontrol mechanism.

1. In practice, theremay be a number of different groups. For example,a group of devices may require a mixed performance requirement amongmaximum throughput, low-latency, and low-power consumption.

JANG ET AL.: VERSATILE ACCESS CONTROL FOR MASSIVE IOT: THROUGHPUT, LATENCY, AND ENERGY EFFICIENCY 1985


metric such as low-latency (LL) and energy efficiency (EE).Activation of nodes is assumed to be bursty due to an exter-nal event for both groups. In addition, the eNodeB isassumed to adopt a probability-based dynamic access classbarring (ACB) technique as used in the commercial cellularnetwork, i.e., 3GPP LTE system [32], which inherently con-trols a number of simultaneous accesses by using a singlenumber called ACB factor p 2 ð0; 1Þ. Let pLL and pEE denotethe ACB factors for LL access and EE access, respectively. Inthe proposed versatile AC mechanism, we consider follow-ing parameters:

� M: number of available preambles� Q: number of available RAR messages� U : number of available PUSCH resources� N : total number of activated nodes in a group� ni: estimated number of backlogged nodes for the ith

slot� pi: ACB factor for the ith slotThe above parameters will be explained in the next sub-

section in detail. Note that Q depends on PDSCH orPDCCH resources, and U depends on PUSCH resources.Since a single RAR message can deliver a single uplinkresource grant for PUSCH [33], the number of allocablePUSCH resources is limited to

K ¼ minfQ;Ug: (1)

Thus, we only consider the number of available preamblesM and the number of allocable PUSCH resources Kthroughout the paper.

2.1 Overall Random Access Procedure

The overall RA procedure consists of five steps, whichinclude the access check (Step 0) and the typical four-stepcontention-based RA procedure (Step 1-4) [34].

(Step 0) Access check: The eNodeB calculates an ACB fac-tor for a specific purpose based on the estimated number ofbacklogged nodes ni and the number of available resourcesM and K, and then it notifies the ACB factor of the ithPRACH pi 2 ½0; 1� at the beginning of the ith PRACH slot.Then, each activated node generates a random numberq 2 ½0; 1� and compares q with pi. If q � pi, then the nodeattempts an RA on the ith PRACH slot. Otherwise, it defersan RA attempt to the ðiþ 1Þth PRACH slot.

(Step 1) Preamble transmission and detection: Each nodewhich passed the access check at the Step 0 randomlyselects a single preamble among M available preambles,and then sends the selected preamble on the ith PRACHslot. More than one node may select the same preamble dueto the random preamble selection property.

(Step 2) Random access response (RAR): After detecting thepreambles from nodes, the eNodeB sends the RAR mes-sages, each of which conveys identity of detected pream-bles, timing advance information, and an initial uplinkPUSCH resource grant for uplink data transmission.

(Step 3) Uplink data transmission: Using the PUSCHresource indicated via the RAR message at the Step 2, thecorresponding node transmits uplink data which conveys adata packet, a radio resource control (RRC) connectionrequest, a tracking area update, or a scheduling request.

(Step 4) ACK message transmission: If the eNodeB successfullydecodes the data transmitted by a single node, it sends backthe ACK message including the node identity (ID) which isobtained from the decoded data, otherwise the eNodeBsends nothing back.

2.2 Preamble Collision Detection Techniques

As noted before, we consider two preamble collisiondetection (PCD) techniques: conventional PCD (C-PCD)and early PCD (E-PCD) [35]. Fig. 2 compares the two techni-ques. The C-PCD technique is performed based on unsuc-cessful data decoding at Step 3 of the RA procedure. Inother works, if multiple nodes send the same preambleat Step 1 and receive the identical RAR message from theeNodeB at Step 2, then all relevant nodes transmit their owndata via the same uplink resource and the eNodeB may notdecode the data successfully. From this unsuccessful decod-ing, the eNodeB detects the preamble collision. If the pream-ble collisions are detected, the eNodeB does not send theACKmessage to the corresponding devices at Step 4. Nodeswhich do not receive the ACK message try to perform theRA procedure at the next PRACH slot.

On the other hand, the PCD can be performed at Step 1of the RA procedure with the E-PCD techniques whichexploit tagged preambles [30], [31], ID transmission [36], ormachine learning algorithms [37]. If the preamble collisionsare detected with E-PCD techniques at Step 1, the eNodeBdoes not send the RAR message to the corresponding nodesat Step 2. Then, nodes which do not receive the RAR mes-sage defer to attempt the RA to the next PRACH slot with-out transmitting uplink data. It is worth noting that the E-PCD techniques can separate the detected preambles intocollision-free preambles and collied preambles and allocatesthe PUSCH resources only for the collision-free preamblesthrough the RAR messages at Step 2.

2.3 Traffic Model

We assume that each of N nodes in a group is activated atthe time x 2 ð0; TactÞ, which follows Beta distribution with

Fig. 2. Comparison of preamble collision detection (PCD) techniques:(a) conventional PCD technique and (b) early PCD technique.



parameters a ¼ 3 and b ¼ 4 [38]:

fXðxÞ ¼ xa�1ðTact � xÞb�1

ðTactÞaþb�1Bða;bÞ ;

where Bða;bÞ ¼ R 10 xa�1ð1� xÞb�1dx denotes the beta func-tion. It is assumed that there exist Iact slots within timeTact and the expected number of newly activated nodesin the ith RA slot is given by �i ¼ N

R titi�1

fXðxÞdx fori ¼ 1; 2; . . . ; Iact.

3 PRELIMINARY PROBABILITIES ON COLLISION-FREE PREAMBLE AND COLLIDED PREAMBLE

In this section, we first define the following discrete randomvariables (RVs):

� D: number of detected preambles,� S: number of collision-free preambles,� C: number of collided preambles,� R: number of undetected preambles,where D ¼ S þ C and R ¼ M �D. Then, we derive sev-

eral preliminary probabilities related to the above RVs,which are used for designing the versatile AC mechanism.

At Step 1 of the RA procedure, each activated nodewhich passes the access check sends an arbitrary preambleamong M available preambles via PRACH, and then theeNodeB detects preambles. In general, detected preamblesare classified into collision-free preambles and collided pre-ambles. The collision-free preamble indicates the detectedpreamble which is sent by a single node, while the collidedpreamble indicates the detected preamble which is sent bymore than two nodes. However, depending on the PCDtechniques (e.g., E-PCD or C-PCD), the eNodeB can or can-not identify the detected preamble is collision-free or col-lided at Step 1 of the RA procedure.

We first consider the following joint probability thatwhen m nodes simultaneously attempt the RAs, any ðrþ sÞpreambles among M preambles are chosen, and then eachof s preambles among chosen ðrþ sÞ preambles is selected(collision-free) only by a sinlge node, and each of remainingr preambles is not selected by any nodes:

CMrþsðsjmÞ

¼ M

rþ s

� �rþ s

s

� � m

s

� �s!fM � ðrþ sÞgm�s

Mm;

(2)

where Mrþs

� �rþss

� �represents the total number of cases that

chooses ðrþ sÞ preambles amongM, and then chooses s col-lision-free preambles among the chosen ðrþ sÞ preambles

[15]. In addition, the term ms

� �s!ðd� sÞm�s represents the

total number of cases that s nodes among m RA-attempting

nodes transmits an exclusive preamble and the remain-

ing ðm� sÞ nodes chooses one of preambles among

fM � ðrþ sÞg preambles. Then, the denominatorMm repre-

sents the total number of cases that m RA-attempting nodes

select a preamble amongM preambles.Let Ei denote the event that the ith preamble is selected

by at most one node, and then the probability of the eventEi whenm nodes attempt RAs simultaneously, i.e.,m nodespass the access check at Step 0, is given by the sum of the

probability that ith preamble is not selected by any nodesand the probability that ith preamble is selected only by asingle node, i.e.,

PrfEijmg ¼ 1� 1

M

� �m

þ m

1

� � 1

M1� 1

M

� �m�1

; (3)

where 1M denotes the probability that a node selects the ith

preamble amongM possibilities.Based on the inclusion-exclusion principle [39], Pr fE1 [

� � � [ EM jmg is written as

Pr [Mi¼1Eijm

� ¼XMi¼1

ð�1Þiþ1Xij¼0

CMi ðjjmÞ

¼XMi¼1

Xij¼0

ð�1Þiþ1 M

i

� �i

j

� �m

j

� �j!ðM � iÞm�j

Mm;

(4)

which denotes the probability that at least one preambleamongM preambles is selected by at most one node with mRA-attempting nodes. Besides, the complementary proba-

bility to Pr [Mi¼1Eijm

� is given by

FðMjmÞ ¼ 1� Pr [Mi¼1Eijm

� ¼ 1�

XMi¼1

Xij¼0

ð�1Þiþ1 M

i

� �i

j

� �m

j


Mm

¼XMi¼0

Xij¼0

ð�1Þi M

i

� �i

j

� �m

j


Mm;

(5)

which represents the probability that each of all M pream-bles are selected by at least two nodes (collided) withm RA-attempting nodes. As a result, we obtain the joint probabil-ity that s preambles are collision-free among d detected pre-ambles whenm nodes simultaneously attempt the RA:

PrfD ¼ d; S ¼ sjmg ¼ CMrþsðsjmÞFðM � r� sjm� sÞ

¼ CMM�dþsðsjmÞFðd� sjm� sÞ

¼Xd�s

i¼0

Xij¼0

M

d

� �d

s

� �m

s

� �s!ðd� sÞm�s

Mm�

ð�1Þi d� s

i

� �i

j

� �m� s

j

� �j!ðd� s� iÞm�s�j

ðd� sÞm�s ;

(6)

where we utilize r ¼ M � d and MM�dþs

� �M�dþs

s

� � ¼ Md

� �ds

� �.

We assume that a Priori distribution of the number ofbacklogged nodes n follows Poisson distribution with meanof n, i.e.,

PðnjnÞ ¼ nn

n!e�n: (7)

It is shown that this assumption significantly reduces thecomputation complexity of designing algorithms [40]. Then,for given ACB factor p and n, the joint probability that s pre-ambles are collision-free among d detected preambles whenm out of n backlogged nodes simultaneously attempt theRA is given by



Prfd; s; n;mjp; ng ¼ Prfd; sjmg|fflfflfflfflfflfflffl{zfflfflfflfflfflfflffl}6

Prfmjp; ng|fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl}BnmðpÞ

Prfnjng|fflfflfflffl{zfflfflfflffl}PðnjnÞ

; (8)

where BnmðpÞ ¼ n

m

� �pmð1� pÞðn�mÞ denotes binomial distri-

bution with ACB factor of p. By marginalizing (8) with

regard to n and m, the joint probability that s preambles are

collision-free among d detected preambles for given p and n

is given by

PrfD ¼ d; S ¼ sjp; ng ¼X1n¼0

Xnm¼0

Prfd; s; n;mjp; ng

¼ M

d

� �d

s

� �e�pn pn

M

� �s�1� pn

Mþ e

pnM

� �d�s

:

(9)

Then, by marginalizing PrfD ¼ d; S ¼ sjp; ng with regard to

S and D, respectively, PrfD ¼ djp; ng and PrfS ¼ sjp; ng are

also given by

PrfD ¼ djp; ng ¼Xds¼0

PrfD ¼ d; S ¼ sjp; ng

¼ M

d

� �e�pn

Xds¼0

d

s

� �pn

M

� �s�1� pn

Mþ e

pnM

� �d�s

¼ M

d

� �e�pn �1þ e

pnM

� �d;

(10)

and

PrfS ¼ sjp; ng ¼XMd¼0

PrfD ¼ d; S ¼ sjp; ng

¼ e�pn pn

M

� �sXMd¼0

M

d

� �d

s

� ��1� pn

Mþ e

pnM

� �d�s

¼ M

s

� �e�pn pn

M

� �sXMd¼0

M � s

d� s

� ��1� pn

Mþ e

pnM

� �d�s

¼ M

s

� �e�pn pn

M

� �s� pn

Mþ e

pnM

� �M�s

;

(11)

respectively. Since C ¼ D� S, we have PrfC ¼ c; S ¼sjp; ng by substituting sþ c for d in Eq. (9). Thus, the proba-

bility that the number of collided preambles is equal to c is

given by

PrfC ¼ cjp; ng ¼XM�c

s¼0

PrfC ¼ c; S ¼ sjp; ng

¼ M

c

� �e�pn �1� pn

Mþ e

pnM

� �c1þ pn

M

� �M�c

:

(12)

As a result, for given p and n, the expected values of RVs D,

S and C are given by

E½Djp; n� ¼XMd¼0

dM

d

� �e�pn �1þ e

pnM

� �d

¼ Me�pn �1þ epnM

� �XMd¼0

M � 1

d� 1

� ��1þ e

pnM

� �d�1

¼ Me�pn �1þ epnM

� �epnMðM�1Þ

¼ M 1� e�pnM

� �;

(13)

E½Sjp; n� ¼XMs¼0

sM

s

� �pn

M

� �se�pn � pn

Mþ e

pnM

� �M�s

¼ pne�pnXMs¼0

M � 1

s� 1

� �pn

M

� �s�1� pn

Mþ e

pnM

� �M�s

¼ pne�pnepnMðM�1Þ

¼ pne�pnM;

(14)

and

E½Cjp; n� ¼ E½Djp; n� � E½Sjp; n�¼ M 1� e�

pnM

� �� pne�

pnM

¼ M �1� pn

Mþ e

pnM

� �e�

pnM;

(15)

respectively.

4 VERSATILE ACCESS CONTROL MECHANISMS

In this section, we propose two representative AC mecha-nisms: low-latency AC mechanism and energy-efficientAC mechanism. For each mechanism, we derive the ACBfactors by considering both the C-PCD technique and theE-PCD technique. Here, we assume that the estimated num-ber of backlogged nodes is equal to n. The estimation algo-rithm for the number of backlogged nodes will be explainedin the next section.

4.1 Low-Latency Access Control

The low-latency requirement can be satisfied by maximiz-ing the access throughput in each slot. Thus, we derive theACB factor for maximizing the access throughput which isdefined as the number of nodes that succeed in the RA.

4.1.1 With C-PCD Technique

Basically, the RA of a certain node is successful when itsends an exclusive preamble and receives an exclusivePUSCH resource grant. Let A denote the access throughput,i.e., the number of nodes that succeed in the RA. Recall thatK denotes the number of allocable PUSCH resources atStep 2 of the RA procedure and D denotes the number ofdetected preambles. By using the well-known total proba-bility theorem, the probability that A ¼ a can be obtained as

PrfA ¼ ag ¼ PrfA ¼ a;D � Kg þ PrfA ¼ a;D > Kg¼ PrfS ¼ a;D � Kg þ PrfS � a;D > KgBs

aKD

� �;

(16)

where Bsa

KD

� � ¼ sa

� �KD

� �a1� K

D

� �s�aand K

D < 1. The termKD < 1 can be regarded as the resource scheduling probabil-ity. Then, the conditional expected value of A for given pand n is given by

E½Ajp; n� ¼XKa¼0

XKd¼0

aPrfD ¼ d; S ¼ ajp; ngþ

XKa¼0

XMd¼Kþ1

Xds¼a

aPrfD ¼ d; S ¼ sjp; ngBsa

KD

� �:

(17)



Then, we obtain the optimal ACB factor p to maximize theaverage access throughput as follows:

pLLconv ¼ argmax0�p�1

E½Ajp; n�: (18)

In particular, whenK ¼ M, Eq. (17) can be rewritten by

E½Ajp; n� ¼XMa¼0

XMd¼0

aPrfD ¼ d; S ¼ ajp; ng

¼ E½Sjp; n�:(19)

Thus, in this case, the optimal pLLconv is found by taking deriv-

ative of E½Sjp; n� ¼ pne�pnM with respect to p, i.e.,

pLLconv ¼M

n: (20)

WhenK < M, on the other hand, it is difficult to find theclosed-form expression for the optimal ACB factor and thuswe approximate E½Ajp; n� as

E½Ajp; n� E½Sjp; n� �min 1;K

E½Djp; n��

; (21)

where we have two components: E½Sjp; n� represents theaverage number of nodes that succeed in preamble trans-

missions, and min 1; KE½Djp;n�

n orepresents the successful

resource scheduling probability. In case that E½DjpLLconv ¼Mn; n� � K, the optimal ACB factor can be approximated as

Eq. (20) since minf1; KE½DjpLLconv;n�

g ¼ 1 and E½Ajp; n� E½Sjp; n�.On the other hand, when E½DjpLLconv ¼ M

n; n� > K, E½Ajp; n�

increases until the ACB factor p approaches to ~pLLconv such

that E½Dj~pLLconv; n� ¼ K and E½Ajp; n� becomes decreased if

p > ~pLLconv since E½Sjp; n� < E½Djp; n� and E½Djp; n� ¼ M

1� e�pnM

� �is an increasing function of p. It implies that the

eNodeB should control the ACB factor p so that it allocates

PUSCH resources to all detected preambles in order to max-

imize access throughput. In other words, the resourcescheduling probability needs to approach to 1 for maximiz-

ing the access throughput.

As a result, we obtain the following ACB factor by solv-

ing E½Djp; n� ¼ M 1� e�pnM

� �¼ K:

~pLLconv ¼ � ln 1� K

M

� �M

n; K < M: (22)

4.1.2 With E-PCD Technique

With E-PCD techniques, the eNodeB allocates resourcesonly for the collision-free preambles, and thus the averageaccess throughput is given by


minða;KÞPrfS ¼ ajp; ng; (23)

where S denotes the number of collision-free preambles.Then, we obtain the optimal ACB factor p to maximize theaverage access throughput as follows:

pLLearly ¼ argmax0�p�1

E½Ajp; n�: (24)

Similar to the case of C-PCD technique, whenK ¼ M,


aPrfS ¼ ajp; ng ¼ E½Sjp; n�; (25)

and thus the optimal pLLearly is also found by

pLLearly ¼M

n: (26)

Note that, when K ¼ M, the optimal ACB factors with theC-PCD and the E-PCD are identical, i.e., pLLconv ¼ pLLearly ¼ M

n.

WhenK < M, on the other hand, it is difficult to find theclosed-form expression for the optimal ACB factor, andthus, we approximate E½Ajp; n� as

E½Ajp; n� E½Sjp; n� �min 1;K

E½Sjp; n��

; (27)

where we have two components: E½Sjp; n� represents theaverage number of nodes that succeed in preamble trans-

missions, and min 1; KE½Sjp;n�

n orepresents the successful

resource scheduling probability. Note that Eq. (27) is differ-

ent from Eq. (21), which comes from the fact that the E-PCD

technique allocates resources only for the collision-free

preambles. In case that E½SjpLLearly ¼ Mn; n� � K, the optimal

ACB factor can be approximated as Eq. (26) since

minf1; KE½SjpLL

early;n�g ¼ 1 and E½Ajp; n� E½Sjp; n�. However,

when E½SjpLLearly ¼ Mn; n� > K, E½Ajp; n� increases until the

ACB factor p approaches to ~pLLearly such that E½Sj~pLLearly; n� ¼ K

and E½Ajp; n� K if ~pLLearly < p � pLLearly. It implies that the

eNodeB cannot allocate resources even for some collision-

free preambles when p > ~pLLearly, which is called resource

allocation failure in this paper. The average number of colli-

sion-free preambles that experience the resource allocation

failure is given by

F ðp; nÞ ¼ 0; if p � ~pLLearly;

E½Sjp; n� �K; if ~pLLearly < p < pLLearly:

(

We obtain the ACB factor p such that E½Sjp; n� ¼ K to maxi-

mize the access throughput and minimize the number of

resource allocation failures (the number of unnecessary

accesses). As a result, we obtain the following ACB factorby solving E½Sjp; n� ¼ pne�

pnM ¼ K:

~pLLearly ¼ �W0 �K

M

� �M

n; K < M; (28)

where W0ðxÞ represents the principle (upper) branch

Lambert W function withW0ðxÞ > �1 [41].



However, if ~pLLearly is used as the ACB factor, the averageaccess throughput becomes slightly lower than K becausethe transmissions occur in a probabilistic manner. To vali-date this, let p denote the ACB factor and BN

n ðpÞ ¼Nn

� �ðpÞnð1� pÞðN�nÞ denotes the probability that n nodes

attempt RAs among N backlogged nodes with the ACBfactor of p. Then, the average number of successfulRA-attempts is given by

Sðp;KÞ ¼XNn¼0

min

"n 1� 1

M

� �n�1; K

#BNn ðpÞ; (29)

where Sð~pLLearly; KÞ is smaller than SðpLLearly; KÞ.On the other hand, let us define unsuccessful RA-attempts

as the attempts either with collided preambles or with colli-sion-free preambles experiencing the resource allocationfailure. The average number of unsuccessful RA-attempts isgiven by

Uðp;KÞ ¼XNn¼0

nBNn ðpÞ � Sðp;KÞ: (30)

Note that the unsuccessful RA-attempt causes energy waste

of nodes. Hence, the ACB factor needs to be carefully con-

trolled between ~pLLearly and pLLearly by considering both Sðp;KÞand Uðp;KÞ in practice. We will further discuss a detailed

setting for p in the section of numerical results.

4.2 Energy Efficient Access Control

Energy efficiency (EE) is defined as the ratio of energy con-sumed for the successful uplink transmissions to energyconsumed for all uplink transmissions during the RA proce-dure at a node.

4.2.1 With C-PCD Technique

The EE with C-PCD technique is given by

E½hEEconvðpÞ� ¼ð1þ xÞE½Sjp; n�min 1;E K

D jp; n �� pnþ xpn min 1;E K

D jp; n �� ¼ ð1þ xÞE½Sjp; n� min 1;E K

D jp; n �� pn 1þ x min 1;E K

D jp; n �� ;

(31)

where x denotes the ratio of energy consumed for the pre-amble transmission at Step 1 over energy consumed for thedata transmission at Step 3 at a node. If we control the ACBfactor such that all nodes with detected preambles receive

resource grants, i.e., E½Djp; n� � K, min 1;E KD jp; n �� ¼ 1,

then E½hEEconvðpÞ� is rewritten as

E½hEEconvðpÞ� ¼ð1þ xÞE½Sjp; n�

1þ xð Þpn ¼ pne�pnM

pn¼ e�

pnM: (32)

The EE with C-PCD technique is a decreasing function of p,

and thus there exists a trade-off between the average access

throughput and the energy efficiency. It is worth noting that

(32) does not depend on the energy ratio of preamble over

data transmissions x. In particular, if pLLconv and ~pLLconv are

used for the ACB factor in (32), then we have

E½hEEconvðpLLconvÞ� ¼ e�1; (33)

and

E½hEEconvð~pLLconvÞ� ¼ 1� K

M; (34)

respectively.

For achieving a certain EE of u for a group of nodes, theACB factor is controlled as:

pEE:uconv ¼ � lnðuÞMn; (35)

which satisfies

E½hEEconvðpÞ� ¼ e�pnM ¼ u:

For energy-efficient AC, the target value of u needs to be

higher than both E½hEEconvðpLLconvÞ� and E½hEE

convð~pLLconvÞ�, which is

set to

u >e�1; if E½DjpLLconv ¼ M

n; n� � K;

1� KM ; if E½DjpLLconv ¼ M

n; n� > K:

�

4.2.2 With E-PCD Technique

The EE with E-PCD technique is given by

E½hEEearlyðpÞ� ¼ð1þ xÞE½Sjp; n�min 1;E K

S jp; n ��

pnþ xE½Sjp; n�min 1;E KS jp; n �� : (36)

If we control the ACB factor such that all nodes with colli-sion-free preambles receive resource grants, i.e., E½Sjp; n� �K,min 1;E K

S jp; n �� ¼ 1, then E½hEEearlyðpÞ� is rewritten as

E½hEEearlyðpÞ� ¼ð1þ xÞE½Sjp; n�pnþ xE½Sjp; n� ¼

ð1þ xÞpne�pnM

pnþ xpne�pnM

¼ ð1þ xÞe�pnM

1þ xe�pnM

: (37)

The EE with E-PCD technique is a decreasing function of p,and thus there exists a trade-off between the average access

throughput and the energy efficiency as well. In contrast to

E½hEEconvðpÞ� in (32), the EE with E-PCD technique E½hEEearlyðpÞ�depends on the energy ratio of preamble over data trans-

missions x. In particular, if pLLearly and ~pLLearly are used for the

ACB factor in (37), then we have

EhhEEearlyðpLLearlyÞ

i¼ ð1þ xÞe�1

1þ xe�1; (38)

and

EhhEEearlyð~pLLearlyÞ

i¼ ð1þ xÞeW0 �K

Mð Þ1þ xeW0 �K

Mð Þ ; (39)

respectively.

For achieving a certain EE of u for a group of nodes, theACB factor is controlled as:

pEE:uearly ¼ � lnu

1þ ð1� uÞx� �

M

n; (40)



which satisfies

E½hEEearlyðpÞ� ¼ð1þ xÞe�pn

M

1þ xe�pnM

¼ u:

For energy-efficient AC, the target value of u needs to be

higher than both E½hEEearlyðpLLearlyÞ� and E½hEE

earlyð~pLLearlyÞ�, which is

set to

u >

ð1þxÞe�1

1þxe�1 ; if E½SjpLLearly ¼ Mn; n� � K;

ð1þxÞeW0 �KMð Þ

1þxeW0 �K

Mð Þ ; if E½SjpLLearly ¼ Mn; n� > K:

8><>:

4.3 Comprehensive Access Control

In Sections 4.1 and 4.2, we proposed low-latency AC andenergy-efficient AC mechanisms, respectively. However, inpractical situation, the eNodeB is required to supportaccesses both from nodes with low-latency QoS and fromnodes with energy-efficient QoS at the same time. First, weassume that eNodeB can identify the access ratio betweenlow-latency access and energy-efficient access. Let g andð1� gÞ denote the access ratios of the low-latency accessand the energy-efficient access, respectively. Then, whenthere exists Nt backlogged nodes at time slot t in total, gNt

and ð1� gÞNt backlogged nodes attempt random accesswith low-latency QoS and energy-efficient QoS, respec-tively, on the average. In the proposed comprehensiveaccess control mechanism, the eNodeB broadcasts tworespective ACB factors pLL and pEE:u, and each node utilizesone of ACB factors depending on its QoS (low-latency orenergy-efficiency) requirement. Consequently, we have thefollowing equation:

gNtpLL þ ð1� gÞNtp

EE:u ¼ Ntpcomp:; (41)

where pcomp: ¼ gpLL þ ð1� gÞpEE:u denotes a comprehensive

ACB factor, which is utilized to estimate Nt as n. Based on

the estimated backlogged size n, ACB factors pLL and pEE:u

are calculated based on the proposed low-latency AC in

(20), (22), (26), and (28) and energy-efficient ACmechanisms

in (35) and (40), respectively.

5 ESTIMATION ON THE NUMBER OF

BACKLOGGED NODES

We extend the pseudo-Bayesian backlog estimation scheme[18] in order to estimate the number of backlogged nodesfor the proposed versatile AC mechanism. The pseudo-Bayesian backlog estimation scheme operates based onthe number of available preambles M and the number ofundetected (idle) preambles r, which can be exactly iden-tified during the preamble detection procedure at Step 1of the RA procedure for both C-PCD and E-PCD techni-ques. Algorithm 1 summarizes the estimation and updatealgorithm of n. The eNodeB first estimates n0 as thenumber of preambles M. Then, the estimated value isupdated as n ¼ nþ Dn. The estimation offset Dn > 0 isgiven by [18]

DnðpÞ ¼ E½njr; p; n� � n ¼ npe�

pnM � r

M

1� e�pnM

!: (42)

For various ACB factors, (42) is rewritten as:

DnðpLLconvÞ ¼ DnðpLLearlyÞ ¼Me�1 � r

1� e�1; (43)

Dnð~pLLconvÞ ¼ � ln 1� K

M

� �M

K

� �ðM �K � rÞ; (44)

Dnð~pLLearlyÞ ¼ �W0 �K

M

� �M

eW0 �KMð Þ � r

M

1� eW0 �KMð Þ

!; (45)

DnðpEE:uconvÞ ¼ � lnðuÞM u � rM

1� u

� �; (46)

DnðpEE:uearlyÞ ¼ � lnð�ÞMu

1þð1�uÞx � rM

1� u1þð1�uÞx

!: (47)

For the comprehensive AC mechanism, the estimationoffset is calculated by Dnðpcomp:Þ, where pcomp: ¼ gpLLþð1� gÞpEE:u. As in [18], we introduce a boosting factor ki inorder to reflect characteristics of the bursty traffic. It impliesthat when we observe that the traffic continuously increasesbased on Dn > 0, we boost the estimation n. At the last line,the new estimation of ni is calculated by ni ¼ ni�1 � ci�1,where ci�1 denotes the number of successful accesses on theði� 1Þth PRACH slot.

Algorithm 1. Estimation and Update Algorithm of n

1: Initialize n0 ¼ M, k0 ¼ 0, and p0.2: Count the number of undetected preambles ri�1

3: Calculate Dn according to (42) with ri�1, pi�1, and ni�1

4: Update ni�1 ¼ ni�1 þ Dn

5: if Dn > 0 then6: ki ¼ ki�1 þ 1 and ni�1 ¼ ni�1 þ ki � Dn " Boosting7: else8: ki ¼ 0 and ni�1 ¼ ni�1

9: end if10: ni ¼ maxðM; ni�1 � ci�1Þ " New estimation of n for slot i

6 NUMERICAL RESULTS

Table 1 summarizes parameters used for computer simula-tions. We assume that random access opportunity (PRACH)appears every 1 ms, i.e., Tinterval ¼ 1 ms, and the number ofpreambles is equal to M ¼ 64.2 The number of allocableresources K is assumed to vary from 10 to 64 and eachgroup of nodes consists of 10000 nodes which try to connectto the network through the RA procedure. The proposedlow-latency AC mechanism aims at the lowest access delayor equivalently the maximum access throughput. For theproposed energy-efficient AC mechanism, the target EEvalue is assumed to be u. In particular, for the

2. The proposed versatile AC mechanism can dynamically adjustACB factors according to the number of preambles, but we focus on thecase of the fixed number of preambles in this paper.



comprehensive AC mechanism, we set the access ratio g to0.5. We evaluate the performances of the proposed ACmechanisms in terms of access throughput, average accessdelay, and energy efficiency. The access delay is defined asthe time difference between the access completion time andthe activation time of a node. For comparison, the perfor-mance of the conventional low-latency AC mechanisms[12], [13], [14], [15], [16], [17], [18] with the ACB factor ofp ¼ M

nis also illustrated. Among [12], [13], [14], [15], [16],

[17], [18], the backlog estimation scheme of [18] is utilized toobtain n representatively.

6.1 With C-PCD Technique

First of all, we evaluate the performance of backlog estimation.Fig. 3 shows the pseudo-Bayesian estimation result for the

number of backlogged nodes in case of energy-efficient ACand low-latency AC, respectively, with K ¼ 64. We observedthat there exist a gap between the estimation and actual val-ues, especially during the bursty period. However, the estima-tion algorithm keeps track of the actual values quite wellduring the traffic descent period. In particular, the backlogestimation under the low-latency AC is quite accurate.

Fig. 4 shows (a) access throughput, (b) average accessdelay, and (c) energy efficiency of the proposed low-latencyAC mechanism with the C-PCD technique. The proposedlow-latency AC mechanism achieves higher access through-put and energy efficiency, and lower average access delaythan those of the conventional low-latency AC mechanismwhen K � 42 since it optimizes the ACB factor by consider-ing not only M but also K, while the conventional ACmechanism takes into account only M. According to theresult, both the proposed and the conventional low-latencyAC mechanism results in the same performances whenthe number of allocable PUSCH resources is sufficientlylarge. The performance gap between two mechanisms is sig-nificant especially when K is small. For example, the aver-age access delay of the proposed low-latency ACmechanism is lower than 200 ms, while the average accessdelay of the conventional low-latency AC mechanism ishigher than 300 ms when K ¼ 18. Furthermore, the EE ofthe proposed low-latency mechanism (0.80) is approxi-mately 4 times higher than the EE of the conventional low-latency AC mechanism (0.21) when K ¼ 10. According tosimulation statistics even though they are not shown in thefigure, when K ¼ 10, the proposed low-latency AC mecha-nism spends 10983 PUSCH resources and 983 PUSCHresources are wasted due to preamble collisions, while theconventional low-latency AC mechanism spends 16738PUSCH resources and 6738 PUSCH resources are wasteddue to preamble collisions for accommodating 10000 nodes.

Fig. 5 shows (a) access throughput, (b) average accessdelay, and (c) energy efficiency of the proposed energy-efficient AC mechanism with the C-PCD technique for vari-ous target EEs (u ¼ 0:9; 0:8; and 0.6). As shown in Fig. 5c, theproposed energy-efficient AC mechanism satisfies the targetEE regardless of K at the cost of increase of average accessdelay as shown in Fig. 5b. As expected, there exists a tradeoffrelationship between the EE and throughput/delay perform-ances. Note that the conventional ACmechanism only yieldsthe maximum EE of 0.38 when K � 42. For the case that

TABLE 1Simulation Parameters and Values

Parameters Values

The number of preambles,M 64The number of allocable PUSCH resources,K 10 64Energy consumption ratio, x 3PRACH interval, Tinterval 1 msActivation duration, Tact 500 msTotal number of RA-attempting nodes, N 10000Access ratio, g 0.5Target energy efficiency, u 0.6/0.8/0.9

Fig. 3. Estimation of backlogged nodes for N ¼ 10000 machine nodes incase of energy-efficient AC and low-latency AC withK ¼ 64.

Fig. 4. Performances of the proposed low-latency AC mechanism with the C-PCD technique for varyingK.



u ¼ 0:6, the proposed energy-efficient AC mechanismachieves higher EE than the target 0.6 whenK � 26 since theproposed low-latency AC mechanism already achieveshigher EE than 0.6 as shown in Fig. 4c. When u ¼ 0:6 andK � 38, the proposed energy-efficiency AC mechanism out-performs the conventional ACmechanism in terms of all per-formance metrics such as access throughput, average accessdelay, and energy efficiency. In addition, when u ¼ 0:8 andK � 20, the same performance tendency is observed.

Fig. 6 shows cumulative distribution functions (CDFs) of(a) the number of access reattempts and (b) the access delayof the proposed AC mechanisms when K ¼ 20. The pro-posed energy-efficient AC mechanism with the target EE of0.9 and 0.8 controls almost 89 and 78 percent of nodes to suc-ceed in the RA in a single attempt. The average number of

reattempts is equal to 0.12 and 0.27 for given EE targets of0.9 and 0.8, respectively. Note that, with the conventionallow-latency AC mechanism, the average number of reat-tempts is equal to 4.27 and only 21 percent of nodes succeedsin the RA in a single attempt. As shown in Fig. 6b, nodesexperience longer access delay with the proposed energy-efficient AC mechanisms when u ¼ 0:9, compared with boththe proposed low-latency AC mechanism and the conven-tional low-latency ACmechanism. However, if the target EEis reduced from 0.9 to 0.8 with the proposed energy-efficientAC mechanism, then the access delay is significantlyreduced from 1168ms to 528ms at the 90th percentile.

Figs. 7a and 7b show the average access delay and theenergy efficiency of the proposed comprehensive AC mecha-nism with the C-PCD technique. Here, we set the access ratio

Fig. 5. Performances of the proposed energy-efficient AC mechanism with the C-PCD technique for varyingK.

Fig. 6. Cumulative distribution function (CDF) of (a) the number of access reattempts and (b) the access delay in energy-efficient AC and low-latencyAC whenK ¼ 20.

Fig. 7. Performances of the proposed comprehensive AC mechanism with the C-PCD technique for varyingK. The access ratio g is set to 0.5.



g to 0.5, which implies that 5000 nodes belong to the low-latency access group and the other 5000 nodes belong to theenergy-efficient access group. In particular, we set the targetenergy efficiency u to 0.9 for the energy-efficient group. Sincethe low-latency group and the energy-efficient group share allpreamble and PUSCH resources together, the low-latencygroup under the proposed comprehensive AC shows loweraverage access delay than that of the low-latency access only(all 10000 nodes attempt low-latency access) as shown inFig. 4b, and the energy-efficient group under the proposedcomprehensive AC also shows lower average access delaythan that of the energy-efficient access only (all 10000 nodesattempt energy-efficient access) as shown in Fig. 5 (b). InFig. 7b, the energy-efficient group under the proposed com-prehensive AC achieves the EE of around 0.8, which is lowerthan 0.9 of the energy-efficient access only as shown in Fig. 5c,whereas the low-latency group under the proposed compre-hensive AC benefits considerably from the energy-efficientgroup in terms of EE, and thus, achieves higher EE than thatof the low-latency access only as shown in Fig. 4c. Accordingto the performance result, we observe that access delay is sig-nificantly sacrificed in order to achieve a very high energy effi-ciency during bursty traffic period.

6.2 With E-PCD Technique

It is worth noting that there exist no previous ACB tech-nique that exploits the E-PCD techniques. Thus, with theE-PCD technique, we only evaluate the performance of the

proposed AC mechanism by changing its own parametersor compare the proposed low-latency AC mechanism to theproposed energy-efficient AC mechanism.

Fig. 8 shows the number of successful and unsuccessfulattempts of the proposed low-latency AC mechanism with

the E-PCD technique, which are denoted by Sðp;KÞ in (29)

and Uðp;KÞ in (30), respectively, when K ¼ 16 and

N ¼ 300. If p � ~pLLearly, then Sðp;KÞ > Uðp;KÞ. However,

Sðp;KÞ < Uðp;KÞ as p approaches to pLLearly. There exists a

tradeoff relationship between the access throughput andthe EE for varying the ACB factor. Hence, we define a novelACB factor as

p ¼ �pLLearly þ ð1� �Þ~pLLearly; (48)

for 0 � � � 1.Fig. 9 shows (a) access throughput, (b) average access

delay, and (c) energy efficiency of the proposed low-latencyAC mechanism with the E-PCD technique for varying � of theACB factor in (48). When � ¼ 1 and � ¼ 0, the ACB factors

becomes pLLearly in (26) and ~pLLearly in (28), respectively. As �

increases, the EE of the proposed mechanism decreases whilethe access throughput increases and the average access delaydecreases, but they are saturated over some �. All performan-ces become improved asK increases. Furthermore, for a givenK, we optimize � so that the EE and the access throughput ofthe proposed low-latency AC mechanism are maximizedwhile the average access delay is minimized, which is denotedby ��. For example, the access throughput is maximized andthe average access delay is minimized for � 2 ½0:6; 1� and themaximum EE is achieved when � ¼ 0:6 for � 2 ½0:6; 1� whenK ¼ 16. Thus, in this case, �� ¼ 0:6. Then, for a given K, theACB factor in the sense that the maximum EE and the maxi-mum access throughput are achieved while the average accessdelay is minimized is obtained by

�pLLearly ¼ ��pLLearly þ ð1� ��Þ~pLLearly: (49)

Fig. 10 shows (a) access throughput, (b) the averageaccess delay, and (c) the energy efficiency of the proposedlow-latency AC mechanism with the E-PCD technique forvarious ACB factors including pLLearly in (26) when � ¼ 0, ~pLLearlyin (28) when � ¼ 0, and �pLLearly with the best �� in (49). Even

though the proposed low-latency AC mechanism with ~pLLearly

Fig. 8. Average numbers of successful and unsuccessful attempts of theproposed low-latency AC mechanism with the E-PCD technique forvarying ACB factor p.

Fig. 9. Performances of the proposed low-latency AC mechanism with the E-PCD technique for varying � whenK ¼ 10; 16; and 20.



(� ¼ 0) results in a slightly longer average access delay whenK � 24 than other cases, it significantly improves the EE.When K > 24, the proposed low-latency AC mechanism

with the E-PCD technique utilizes only pLLearly since

E½SjpLLearly ¼ Mn; n� � K.

Fig. 11 shows (a) the access throughput, (b) the average

access delay, and (c) the energy efficiency of the proposed

energy-efficient AC mechanism with the E-PCD technique.

As shown in the figure, the proposed energy-efficient ACmechanism effectively satisfies the target EE requirementseven though there exist a little dissatisfaction for some Kdue to approximation error in obtaining the ACB factor.

Fig. 12 shows the CDF of (a) the number of access reat-tempts and (b) the access delay of the proposed energy-effi-cient ACmechanismwith u ¼ 0:9; 0:8 and the proposed low-latency AC mechanism with the best ��. With the proposedenergy-efficient ACmechanism, 63 percent and 50 percent ofnodes succeed the RA in a single attempt and the averagenumber of access reattempts is equal to 0.51 and 0.95 whenu ¼ 0:9 and 0.8, respectively. On the other hand,with the pro-posed low-latency AC mechanism, 35 percent of nodes suc-ceed the RA in a single attempt and the average number ofaccess reattempts is equal to 1.73. However, as shown inFig. 12b, at the 90th percentile, the access delays of the pro-posed energy-efficient AC mechanism with the E-PCD

Fig. 10. Performances of the proposed low-latency AC mechanism with the E-PCD technique for varyingK.

Fig. 11. Performances of the proposed energy-efficient AC mechanism with the E-PCD technique for varyingK.

Fig. 12. Cumulative distribution function (CDF) of (a) the number of access reattempts and (b) the access delay of the proposed ACmechanisms withthe E-PCD technique whenK ¼ 20.



technique is equal to 325 ms and 270 ms for u ¼ 0:9 and 0.8,respectively, while that of the proposed low-latency ACmechanismwith the E-PCD technique is equal to 250ms.

7 CONCLUSION

In this paper, we proposed a versatile access control mecha-nism for massive IoT networks, which effectively satisfiesvarious QoS requirements: low-latency for mission-criticalservices and high energy efficiency (EE) for battery-pow-ered sensor applications. We adopted the early collisiondetection technique as well as the conventional preamblecollision detection technique at the first step of the randomaccess (RA) procedure. For optimizing the access class bar-ring factor of the proposed AC mechanism, we consideredall radio resources which are involved in the overall RAprocedure. The proposed low-latency AC mechanismachieves lowest access delay by aggressively allowing forIoT devices to attempt the RA at the cost of the decreasedEE, while the proposed energy-efficient AC mechanismachieves a target EE by restricting attempts from the IoTdevices at the cost of increased access delay. Compared tothe conventional AC mechanism, the proposed low-latencyAC mechanism further reduces the access delay, whileimproving the EE under diverse resource constraints. Fur-thermore, we proposed a comprehensive access controlmechanism which can simultaneously support two groupsof devices requiring low-latency and ultra-low power con-sumption, respectively.

ACKNOWLEDGMENTS

This work was supported in part by the SUTD-ZJUResearch Collaboration under Grant SUTD- ZJU/RES/01/2016, in part by the SUTD-ZJU Research Collaborationunder Grant SUTD-ZJU/RES/05/2016, in part by the MSIT(Ministry of Science and ICT), Korea, under the ITRC(Infor-mation Technology Research Center) support program(IITP-2019-2017-0-01635) supervised by the IITP(Institutefor Information & communications Technology Promotion),and in part by the National Research Foundation of Korea(NRF) grant funded by the Korea government (MSIT)(NRF-2018R1C1B6008126).

REFERENCES

[1] C. X. Mavromoustakis, G. Mastorakis, and J. M. Batalla, Internet ofThings (IoT) in 5G Mobile Technologies. vol. 8. Berlin, Germany:Springer, 2016.

[2] 3GPP, Study on Scenarios and Requirements for Next GenerationAccess Technologies, 3GPP TR 38.913 V14.3.0, June 2017.

[3] ITU-R, Minimum Requirements Related to Technical Performance forIMT-2020 Radio Interface(s), ITU-R M.2410–0, Nov. 2017.

[4] M. Li, R. Yu, P. Si, and Y. Zhang, “Energy-efficient machine-to-machine (M2M) communications in virtualized cellular net-works with mobile edge computing (MEC),” IEEE Trans. MobileComput., to be published, doi: 10.1109/TMC.2018.2865312.

[5] H. Li, K. Ota, and M. Dong, “Learning IoT in edge: Deep learningfor the internet of thingswith edge computing,” IEEENetw., vol. 32,no. 1, pp. 96–101, Jan./Feb. 2018.

[6] S.-C. Lin and K.-C. Chen, “Cognitive and opportunistic relay forQoS guarantees in machine-to-machine communications,” IEEETrans. Mobile Comput., vol. 15, no. 3, pp. 599–609, Mar. 2016.

[7] J. Huang, C. Xing, S. Y. Shin, F. Hou, and C. Hsu, “OptimizingM2M communications and quality of services in the IoT for sus-tainable smart cities,” IEEE Trans. Sustain. Comput., vol. 3, no. 1,pp. 4–15, Jan.-Mar. 2018.

[8] A. Sivanathan, H. H. Gharakheili, F. Loi, A. Radford, C.Wijenayake,A. Vishwanath, and V. Sivaraman, “Classifying IoT devices in smartenvironments using network traffic characteristics,” IEEE Trans.Mobile Comput., to be published, doi: 10.1109/TMC.2018.2866249.

[9] T. P. de Andrade, C. A. Astudillo, and N. L. da Fonseca,“Allocation of control resources for machine-to-machine andhuman-to-human communications over LTE/LTE-A networks,”IEEE Internet Things J., vol. 3, no. 3, pp. 366–377, Jun. 2016.

[10] T. P. deAndrade, C.A.Astudillo, L. R. Sekijima, andN. L. da Fonseca,“The random access procedure in long term evolution networks forthe internet of things,” IEEECommun.Mag., vol. 55, no. 3, pp. 124–131,Mar. 2017.

[11] N. Xia, H.-H. Chen, and C.-S. Yang, “Radio resource management inmachine-to-machine communications-a survey,” IEEE Commun. Sur-veys Tuts., vol. 20, no. 1, pp. 791–828, Jan.-Mar. 2018.

[12] S. Duan, V. Shah-Mansouri, and V. W. S. Wong, “Dynamic accessclass barring for M2M communications in LTE networks,” in Proc.IEEE GLOBECOM, Dec. 2013, pp. 4747–4752.

[13] Z. Wang and V. W. Wong, “Joint access class barring and timingadvance model for machine-type communications,” in Proc. IEEEInt. Conf. Commun., Jun. 2014, pp. 2357–2362.

[14] A. Laya, L. Alonso, and J. Alonso-Zarate, “Is the random accesschannel of LTE and LTE-A suitable for M2M communications? Asurvey of alternatives,” IEEE Commun. Surveys Tuts., vol. 16, no. 1,pp. 4–16, Jan.-Mar. 2014.

[15] M. Tavana, V. Shah-Mansouri, and V. W. Wong, “Congestion con-trol for bursty M2M traffic in LTE networks,” in Proc. IEEE Int.Conf. Commun., Jun. 2015, pp. 5815–5820.

[16] Z. Wang and V. W. Wong, “Optimal access class barring for sta-tionary machine type communication devices with timing advanceinformation,” IEEE Trans. Wireless Commun., vol. 14, no. 10,pp. 5374–5387, Oct. 2015.

[17] S. Duan, V. Shah-Mansouri, Z. Wang, and V. W. Wong, “D-ACB:Adaptive congestion control algorithm for bursty M2M traffic inLTE networks,” IEEE Trans. Veh. Technol., vol. 65, no. 12,pp. 9847–9861, Dec. 2016.

[18] H. Jin,W. Toor, B. C. Jung, and J. B. Seo, “Recursive Pseudo-Bayesianaccess class barring for M2M communications in LTE systems,”IEEE Trans. Veh. Technol., vol. 66, no. 9, pp. 8595–8599, Sep. 2017.

[19] S. Vural, N. Wang, P. Bucknell, G. Foster, R. Tafazolli, andJ. Muller, “Dynamic preamble subset allocation for RAN slicing in5G networks,” IEEE Access, vol. 6, pp. 13 015–13 032, 2018.

[20] C. Y. Ho and C.-Y. Huang, “Energy-saving massive access controland resource allocation schemes for M2M communications inOFDMA cellular networks,” IEEE Wireless Commun. Lett., vol. 1,no. 3, pp. 209–212, Jun. 2012.

[21] H. S. Dhillon, H. C. Huang, H. Viswanathan, and R. A. Valenzuela,“Power-efficient system design for cellular-based machine-to-machine communications,” IEEE Trans. Wireless Commun.,vol. 12, no. 11, pp. 5740–5753, Nov. 2013.

[22] A. Azari and G. Miao, “Energy efficient mac for cellular-basedM2M communications,” in Proc IEEE Global Conf. Signal andInform. Process., Dec. 2014, pp. 128–132.

[23] G. Miao, A. Azari, and T. Hwang, “E2-MAC: Energy efficientmedium access for massive M2M communications,” IEEE Trans.Commun., vol. 64, no. 11, pp. 4720–4735, Nov. 2016.

[24] A. Aijaz, M. Tshangini, M. R. Nakhai, X. Chu, and A.-H. Aghvami,“Energy-efficient uplink resource allocation in LTE networks withM2M/H2H co-existence under statistical QoS guarantees,” IEEETrans. Commun., vol. 62, no. 7, pp. 2353–2365, Jul. 2014.

[25] M. A. Mehaseb, Y. Gadallah, A. Elhamy, and H. Elhennawy,“Classification of LTE uplink scheduling techniques: An M2Mperspective,” IEEECommun. Surveys Tuts., vol. 18, no. 2, pp. 1310–1335,Apr.-Jun. 2016.

[26] D. Zhang, Z. Zhou, S.Mumtaz, J. Rodriguez, andT. Sato, “One inte-grated energy efficiency proposal for 5G IoT communications,”IEEE Internet Things J., vol. 3, no. 6, pp. 1346–1354, Dec. 2016.

[27] Z. Yang, W. Xu, Y. Pan, C. Pan, and M. Chen, “Energy efficientresource allocation in machine-to-machine communications withmultiple access and energy harvesting for IoT,” IEEE InternetThings J., vol. 5, no. 1, pp. 229–245, Feb. 2018.

[28] A. Azari and G. Miao, “Network lifetime maximization for cellular-basedM2M networks,” IEEE Access, vol. 5, pp. 18 927–18 940, 2017.

[29] N. M. Balasubramanya, L. Lampe, G. Vos, and S. Bennett, “Ontiming reacquisition and enhanced primary synchronization sig-nal (ePSS) design for energy efficient 3GPP LTE MTC,” IEEETrans. Mobile Comput., vol. 16, no. 8, pp. 2292–2305, Aug. 2017.



http://dx.doi.org/10.1109/TMC.2018.2865312

http://dx.doi.org/10.1109/TMC.2018.2866249

[30] H. S. Jang, S. M. Kim, H.-S Park, and D. K. Sung, “An early pream-ble collision detection scheme for cellular M2M random access,”IEEE Trans. Veh. Technol., vol. 66, no. 7, pp. 5974–5984, Jul. 2017.

[31] H. S. Jang, S. M. Kim, H.-S Park, and D. K. Sung, “A preamble col-lision resolution scheme via tagged preambles for cellular IoT/M2M communications,” IEEE Trans. Veh. Technol., vol. 67, no. 2,pp. 1825–1829, Feb. 2018.

[32] 3GPP, Study on RAN Improvements for Machine-Type Communica-tions, 3GPP TR 37.868 V11.0.0, Oct. 2011.

[33] 3GPP, Medium Access Control (MAC) Protocol Specification (Release14), 3GPP TS 36.321 V14.6.0, Mar. 2018.

[34] S. Sesia, I. Toufik, and M. Baker, LTE - The UMTS Long Term Evolu-tion: From Theory to Practice. Hoboken, NJ, USA: Wiley, 2009.

[35] H. S. Jang, B. C. Jung, andD. K. Sung, “Dynamic access control withresource limitation for group paging-based cellular IoT systems,”IEEE Internet Things J., vol. 5, no. 6, pp. 5065–5075, Dec. 2018.

[36] N. Zhang, G. Kang, J. Wang, Y. Guo, and F. Labeau, “Resourceallocation in a new random access for M2M communications,”IEEE Commun. Lett., vol. 19, no. 5, pp. 843–846, May 2015.

[37] D. Magrin, C. Pielli, C. Stefanovic, and M. Zorzi, “Enabling LTERACH collision multiplicity detection via machine learning,”arXiv:1805.11482, 2018.

[38] 3GPP, [70bis#11]-LTE: MTC LTE Simulations, 3GPP 3GPP TSG-RANWG2meeting# 71 R2–104663, Aug. 2010.

[39] W. Feller, An Introduction to Probability Theory and Its Application.Hoboken, NJ, USA: Wiley, 1968.

[40] R. Rivest, “Network control by bayesian broadcast,” IEEE Trans.Inf. Theory, vol. IT-33, no. 3, pp. 323–328, May 1987.

[41] R.M. Corless,G.H.Gonnet, D. E.Hare, D. J. Jeffrey, andD. E. Knuth,“On the LambertW function,” Advances Comput. Math., vol. 5, no. 1,pp. 329–359, 1996.

Han Seung Jang (S’14–M’18) received the BSdegree in electronics and computer engineeringfrom Chonnam National University, Gwangju,Korea, in 2012, and the MS and PhD degrees inelectrical engineering from the Korea AdvancedInstitute for Science and Technology (KAIST),Daejoen, Korea, in 2014 and 2017, respectively.From May 2018 to February 2019, he was a post-doctoral fellow with the Information SystemsTechnology and Design (ISTD) Pillar, SingaporeUniversity of Technology and Design (SUTD),

Singapore. He was also a post-doctoral fellow from September 2017 toApril 2018 with Chungnam National University, Daejoen, Korea. He iscurrently an assistant professor with the School of Electrical, ElectronicCommunication, and Computer Engineering, Chonnam National Univer-sity, Yeosu, Korea. His research interests include cellular Internet-of-things (IoT)/machine-to-machine (M2M) communications, machinelearning, smart grid, and energy ICT. He is a member of the IEEE.

Hu Jin (S’07–M’12–SM’18) received the BEdegree in electronic engineering and informationscience from the University of Science and Tech-nology of China, Hefei, China, in 2004, and theMS and PhD degrees in electrical engineeringfrom the Korea Advanced Institute of Scienceand Technology (KAIST), Daejeon, South Korea,in 2006 and 2011, respectively. From 2011 to2013, he was a post-doctoral fellow with TheUniversity of British Columbia, Vancouver, BC,Canada. From 2013 to 2014, he was a research

professor with Gyeongsang National University, Tongyeong, SouthKorea. Since 2014, he has been with the Division of Electrical Engineer-ing, Hanyang University, Ansan, South Korea, where he is currently anassociate professor. His research interests include medium-access con-trol and radio resource management for random access networks andscheduling systems considering advanced signal processing and queue-ing performance. He is a senior member of the IEEE.

Bang Chul Jung (S’02–M’08–SM’14) receivedthe BS degree in electronics engineering fromAjou University, Suwon, Korea, in 2002, and theMS and PhD degrees in electrical & computerengineering from KAIST, Daejeon, Korea, in2004 and 2008, respectively. He was a seniorresearcher/research professor with the KAISTInstitute for Information Technology Conver-gence, Daejeon, Korea, from January 2009 toFebruary 2010. From March 2010 to August2015, he was a Faculty of Gyeongsang National

University. He is currently an associate professor in the Department ofElectronics Engineering, Chungnam National University, Daejeon,Korea. He has served as associate editor of IEICE Transactions on Fun-damentals of Electronics, Communications, and Computer Sciencessince 2018. His research interests include wireless communication sys-tems, internet-of-things (IoT) communications, statistical signal process-ing, information theory, interference management, random access, radioresource management, cooperative relaying techniques, in-networkcomputation, and mobile computing. He was the recipient of severalawards including the Fifth IEEE Communication Society Asia-PacificOutstanding Young Researcher Award in 2011 and he has beenselected as a winner of Haedong Young Scholar Award in 2015, which issponsored by the Haedong foundation and given by Korea Institute ofCommunication and Information Science (KICS), and the Best PaperAward of KICS Journal in 2018, and several Best Paper Awards of KICSConferences in 2017/2018. He is a senior member of the IEEE.

TonyQ.S.Quek (S’98-M’08-SM’12-F’18) receivedtheBEandMEdegrees in electrical and electronicsengineering from the Tokyo Institute of Technol-ogy, Tokyo, Japan, in 1998 and 2000, respectively,and the PhD degree in electrical engineering andcomputer science from theMassachusetts Instituteof Technology, Cambridge, MA, in 2008. Currently,he is a tenured associate professor with the Singa-pore University of Technology and Design (SUTD).He also serves as the Acting Head of ISTD Pillarand the deputy director of the SUTD-ZJU IDEA.

His current research topics include wireless communications and network-ing, internet-of-things, network intelligence, wireless security, and big dataprocessing. He has been actively involved in organizing and chairing ses-sions, and has served as amember of the technical program committee aswell as symposium chairs in a number of international conferences. He iscurrently an elected member of IEEE Signal Processing Society SPCOMTechnical Committee. He was an executive editorial committee memberand an editor for transactions and journals.. He is a co-author of the booksSmall Cell Networks: Deployment, PHYTechniques, and Resource Alloca-tion and the bookCloudRadio Access Networks: Principles, Technologies,andApplications byCambridgeUniversity Press in 2013 and 2017, respec-tively. Dr. Quek was honored with several awards including the 2017CTTCEarly Achievement Award, the 2017 IEEEComSoc APOutstanding PaperAward, and the 2016-2018 Clarivate Analytics Highly Cited Researcher.He is a distinguished lecturer of the IEEE Communications Society and afellow of the IEEE.

" For more information on this or any other computing topic,please visit our Digital Library at www.computer.org/csdl.



versatile access control for massive iot: throughput, latency,...

Documents