Professional Documents
Culture Documents
Abstract—The massive amounts of machine-type user equip- referred to as orthogonal random access (ORA) scheme in the
arXiv:1705.01235v1 [cs.IT] 3 May 2017
ments (UEs) will be supported in the future fifth generation (5G) following sections). The four messages include Preamble, Ran-
networks. However, the potential large random access (RA) delay dom Access Response (RAR), Initial Layer 3 message (Msg3)
calls for a new RA scheme and for a detailed assessment of
its performance. Motivated by the key idea of non-orthogonal and Contention Resolution (CR) [5]. A periodic sequence
multiple access, the non-orthogonal random access (NORA) of time-frequency resources called random access slots (RA
scheme based on successive interference cancellation (SIC) is slots) are reserved in the Physical Random Access Channel
proposed in this paper to alleviate the access congestion problem. (PRACH) for preamble transmission [6]. Whenever a UE
Specifically, NORA utilizes the difference of time of arrival to triggers the RA procedure, it transmits a preamble randomly
identify multiple UEs with the identical preamble, and enables
power domain multiplexing of collided UEs in the following chosen from the available orthogonal pseudo-random pream-
access process, while the base station performs SIC based on bles periodically broadcast by the eNodeB in the next available
the channel conditions obtained through preamble detection. Our RA slot. There are up to 64 available preambles within each
analysis show that the performance of NORA is superior to the cell [7]. So if more than 64 UEs make RA attempts in one
conventional orthogonal random access (ORA) scheme in terms specific RA slot, the collision is inevitable. The UEs whose
of the preamble collision probability, access success probability
and throughput of random access. Simulation results verify our preambles have been successfully recognized by the eNodeB
analysis and further show that our NORA scheme can improve will receive RAR and transmit Msg3 on the Physical Uplink
the number of the supported UEs by more than 30%. Moreover, Shared Channel (PUSCH). In 3GPP RA operation, collision
the number of preamble transmissions and the access delay for happens if more than one UE select the same preamble. The
successfully accessed UEs are also reduced significantly by using UEs that experience collision will not be scheduled for Msg3
the proposed random access scheme.
transmission, and will return to preamble transmission [8]. As
Index Terms—Random access, collision probability, through- the number of UEs grows, the collision becomes more and
put, access delay.
more frequent and finally leads to congestion. The UEs end up
transmitting preambles repeatedly until the maximum allowed
I. I NTRODUCTION number of preamble transmissions is reached. Then the UEs
parameter is adaptively changed based on the amount of scheme is proposed in [19], which spatially partitions the
available RBs and the traffic load. In [11], two dynamic ACB cell coverage into multiple regions and reduces the cyclic
algorithms for fixed and dynamic preamble allocation schemes shift size to generate more preambles. However, the extent
are proposed to determine the barring factors without priori of the RA resource increase is limited and severe collision
knowledge of the number of MTC devices. [9] formulates and retransmission are still inevitable.
an optimization problem to determine the optimal barring The RAN overload problem can also be tackled by effi-
parameter which maximizes the expected number of MTC ciently utilizing the radio resources for the RA process. In
devices successfully served in each RA slot. [12] proposes a [20], the authors propose a novel random access scheme based
two-stage ACB scheme to increase access success probability. on fixed timing alignment information to reduce collision
In the first stage, the UEs use the barring factor broadcast by probability given a large number of fixed-location machine-
the eNodeB. The UEs which pass the ACB check are viewed to-machine (M2M) devices. In the scheme proposed in [21],
as primary UEs and allowed to select non-special preambles M2M UEs form coalitions and perform relay transmission with
randomly, while the UEs which fail are treated as secondary an objective to reduce network congestion. To increase the
UEs and select the special preambles. In the second stage, each preamble detection probability, the authors of [22] propose
secondary UE calculates its barring probability independently an enhanced RA scheme, in which the eNodeB adopts the
based on the expected number of secondary UEs. In terms transmission time difference to detect the UEs which utilizes
of the backoff factor, [13] compares the performance of the same preamble. Furthermore, the eNodeB creates multiple
uniform backoff (UB) and binary exponential backoff (BEB) RARs in response to detected UEs which select the same
algorithms and proposes a new algorithm to adaptively adjust preamble. However, this scheme fails to consider the limi-
the backoff window size under unsaturated traffic conditions. tation on PUSCH resources. With the increase of successfully
EAB extends the granularity of the access class to distin- transmitted preambles, the limited PUSCH resources may be
guish multiple classes, which has been introduced in 3GPP another bottleneck for RA process.
standard to throttle the access of Machine Type Communica- In summary, most of existing solutions to improve LTE RA
tion (MTC) devices [17]. A prioritized random access with performance mainly focus on controlling traffic or increasing
dynamic access barring (PRADA) framework is proposed in available RA resources. However, the available RA resources
[14], which optimizes the EAB parameters such as activation are limited. Moreover, the access delay cannot be guaranteed
time, barring opportunity, and backoff time. PRADA includes by controlling traffic.
two components: pre-allocation of different amount of RA Successive interference cancellation (SIC) enables through-
slots for different classes and dynamic access barring (DAB). put efficiency enhancements by utilizing collided packets for
The average number of successful preambles is observed in decoding instead of discarding them [23], which makes it
each two neighboring RA slots to estimate the RA load. an obvious candidate for alleviating RA congestion problem.
If the RA load is heavy, EAB is triggered and the RA [24] evaluates the random access throughput performance of
attempts of the UEs with their first preamble transmissions asynchronous code division multiple access (CDMA) systems
are deferred for a long time. However, there are concerns that with interference cancellation receivers. Power randomiza-
EAB may be frequently activated and deactivated since the tion is explored to aid iterative receiver processing. [25]
number of successful preambles drastically varies for bursty proposes a code-division random access (CDRA) scheme,
RA arrivals. It will result in performance deterioration, for which adopts specific sets of codewords to spread the uplink
instance, decrease in RA success probability. resources in a non-orthogonal manner among users. The
To guarantee throughput, [15] proposes a game-theoretic codewords assignment scheme allows a random subset of
framework to dynamically allocate additional RA resources to users communicating single bits to the base station (BS).
MTC devices. [16] elaborates a MTC specific backoff scheme It is worth noting that CDRA uses a convex optimization-
which introduces a dynamic backoff indicator assignment based multiple user detection algorithm to avoid obtaining
algorithm to reduce RA collision probability. The pull-based the delays and channel state information (CSI) of the users
scheme [17] allows the eNodeB to control network load by at the BS. Moreover, [23, 26, 27] employ SIC to resolve
dominating the paging operation, where MTC devices will collisions in the tree (also known as splitting) algorithm to
trigger RA process upon receiving a paging message. improve random access throughput. These schemes rely on the
The main idea behind aforementioned schemes is to disperse property of tree algorithm where all packets are retained one-
the transmission of the access request to control overload and by-one in line with the underlying tree structure. However, this
increase the access probability within a relatively short time. property is not available in LTE RA assumption. There also
In spite that these approaches can reduce access collision to a has been extensive studies on random access networks with
certain degree, the retransmission of numerous UEs can again SIC-enabled multi-packet reception capabilities. Nevertheless,
aggravate collision and further increase the access delay. most of recent works focus on the theoretical analysis of
Moreover, several schemes have been developed to mitigate capture probability [28] and power allocation scheme design
collision and reduce access delay by increasing available [29, 30]. Few studies have considered the backoff and re-
RA resources. For example, Thomsen et al. [18] proposes transmission process in LTE RA procedure. To the best of
a code-expanded RA scheme which adopts the concept of our knowledge, systematic performance analysis of the SIC-
access codeword to increase the amount of available contention enabled RA process under LTE scenario has not been reported
resources in the RA process. Moreover, a preamble reuse yet. Recently, non-orthogonal multiple access (NOMA) has
3
received many interests [31–34]. Combined with SIC, NOMA Moreover, the average number of preamble transmissions
allows simultaneously transmissions of multiple UEs with and average access delay are reduced remarkably, and are
different powers [32]. Nonetheless, recent studies on NOMA only half as much as those of the ORA scheme in the
mainly focus on performance analysis of data transmission best case. Comparisons with state-of-the art EAB scheme
process. In practical networks such as LTE-A and future 5G further validate the superiority of the NORA scheme.
networks, the introduction and realization of NOMA and SIC The rest of the paper is organized as follows. A detailed
in the random access process could be very challenging. As description of the NORA mechanism is presented in Section
far as we’re concerned, corresponding protocol and mechanism II. Section III depicts the analytical model of the proposed
design which is compatible with existing LTE standard has not NORA procedure. The performance metrics and simulation
been reported yet. analysis are given in Section IV, while Section V concludes
the paper.
B. Motivation and Contribution
Massive-connections, high-reliability and low-latency are II. N ON - ORTHOGONAL RANDOM ACCESS MECHANISM
typical technical scenarios for Internet of Things (IoT) in
the 5G network. Specifically, smartphones, tablets and M2M In this section, we give a detailed description of the NORA
communications such as environmental monitoring and smart scheme, which consists of PRACH preamble transmission,
meter reading applications have generated an explosive growth random access response, initial layer 3 message transmission
in the number of UEs. The surging RA attempts of enormous and contention resolution (as illustrated in Fig. 1).
UEs call for more efficient and robust RA mechanisms. Moti-
vated by the idea of NOMA and SIC, we propose a SIC-based A. Preamble Transmission
non-orthogonal random access (NORA) mechanism which is Each UE first receives the system information broadcast on
easily applicable to existing LTE standard and future 5G Physical Broadcast Channel (PBCH) and acquires necessary
standards. In contrast to ORA scheme, NORA facilitates the configuration information to complete the RA process [35].
simultaneous transmission of Msg3 of collided UEs instead of The information includes PRACH configuration information
conducting retransmission of preambles, which avoids further such as PRACH Configuration Index, PRACH Frequency
exacerbating collision without increasing demands on PUSCH Offset, Root Sequence Index, etc. and RACH configuration
resources. The information of UE locations and channel con- information such as Number of RA Preambles, Maximum
ditions is utilized to realize power domain multiplexing on Number of Preamble Transmission, RA Response Window
the UE side and SIC on the base station side. Hereafter, we Size, Power Back-off Offset, MAC Contention Resolution
investigate the performance of NORA in terms of through- Timer, etc. When a UE starts to perform random access,
put, preamble collision probability, access success probability, it randomly selects a preamble sequence from the available
access delay, and the number of preamble transmissions. preambles broadcast by the base station and transmits it in
Simulation results show that our NORA scheme can improve the next available RA slot. Preamble sequences are identi-
the number of the supported UEs by more than 30%. The fied by their Random Access Preamble Identity (RAPID).
number of preamble transmissions and the access delay for There is also a one-to-one mapping between Random Ac-
successfully accessed UEs are also reduced remarkably. cess Radio Network Temporary Identifier (RA-RNTI) and the
The contributions of this paper are summarized as follows: time/frequency resources used by the PRACH preamble.
• We propose a novel NORA mechanism which utilizes
the spatial distribution characteristic of UEs. NORA
integrates the arrival time-based multi-preamble detection B. Preamble Detection and RAR transmission
and distance-based RAR reception schemes to effectively 1) Arrival time based multi-preamble detection: The base
improve the preamble transmission success probability. station first extracts the relevant PRACH signals within
The corresponding RAR message format is tailored for specific time/frequency resources through time-domain sam-
practical realizations. pling and frequency-tone extraction. Then the base station
• We derive an analytical model to investigate the tran- computes the PRACH preamble power delay profile (PDP)
sient behavior of the NORA process with non-stationary through frequency-domain periodic correlation. Since different
arrivals. Realistic assumptions, such as UEs’ positions PRACH preambles are generated from cyclic shifts of a
and channel conditions, are considered to show practical common root sequence, the periodic correlation operation
results. provides in one shot the concatenated PDPs of all preambles
• We provide a comprehensive analysis of the NORA per- derived from the same root sequence, as shown in Fig. 2.
formance, including throughput of preamble transmission Each cyclic shift defines a Zero Correlation Zone (ZCZ),
and random access, collision/access success probability, i.e. detection zone for corresponding preamble. The preamble
the cumulative distribution function (CDF) of the num- detection process consists of searching the PDP peaks above
ber of preamble transmissions and access delay for the a detection threshold within each ZCZ. The length of each
successfully accessed UEs, average number of preamble ZCZ is determined by the cell size. When the cell size is
transmissions and average access delay. more than twice the distance corresponding to the maximum
• With the proposed NORA scheme, the throughput of delay spread, the base station may be able to differentiate
random access process is increased by more than 30%. the PRACH transmissions of two UEs which select the same
4
UE Base station
ķ Preamble Transmission
Preamble format System Information
ĸ Preamble Detection and RAR Transmission
CP Preamble PBCH
PRACH Preamble
Ĺ Initial layer 3 message transmission
ķ PRACH ĺ User Separation and Contention Resolution
Preamble DCI message format 1
Detection RA-RNTI DL RBs for PDSCH
User Traffic
PUSCH
loss of 0dB. MU denotes the number of available resource mitting the initial layer 3 message. If the UE does not receive a
blocks in UL Grant while P L denotes the downlink path response from the base station before the timer expires, it will
loss estimate. α represents the reduced rate of transmit power return to the procedure of transmitting PRACH preambles.
increase due to fractional power control [36].
In NORA, the base station performs user separation based
on SIC, which requires diverse arrived power of UEs. To
III. A NALYTICAL MODEL FOR NORA
obtain this, the power back-off scheme [33] is introduced. The
transmit power of the i-th UE in a NORA group is expressed
by The parameters of the NORA process considered in this
paper are summarized in Table I. U depicts the total number
PU,i = min {PUmax , PO U −(i−1) δ+10log10 (MU,i )+αP Li } ,of UEs in the cell. Assume that all UEs attempt to access the
(2) network over a period of time TAP with the arrival distribution
where δ is the Power Back-off Offset broadcast on PBCH. p(t) [17].
Eq. (2) implies that the received power of UEs in a NORA We evaluate the performance of the RA procedure within
group gradually degrades with a step of δ, which is beneficial the time interval TRAI , which ranges from the first preamble
to cancel the co-channel interference successively. Since the transmission of the first attempted UE to the completion of
case that PU,i is the same as PUmax only occurs in extreme the RA procedure of the last attempted UE. TRAI is defined
cases, which are not the focus of this paper, the transmit power as [4]
of the i-th UE in a NORA group is assumed to be TRAI = TAP + TW + TRA , (4)
PU,i = PO U − (i − 1) δ + 10log10 (MU,i ) + αP Li . (3) in which TAP denotes the arrival period, TW represents the
The power back-off order of UEs in a NORA group is average time used by the last UE to wait for the next available
decided according to the TA value. The UE with a larger RA slot (unit: sub-frame) and TRA denotes the maximum
TA will be assigned a larger order i, which indicates that the period of time required by the last UE to complete the RA
corresponding received power PU,i is smaller. procedure.
2) Initial layer 3 message transmission with power domain Represent TRAI in terms
l of mthe number of RA slots K, then
multiplexing: After power back-off and timing alignment with K is defined as K = TTRAP RAI
, where TRAP is the RA slot
base station, the UEs in a NORA group will transmit their period.
respective initial layer 3 messages within the same RBs of
PUSCH. The initial layer 3 message contains information of
UE’s Temp C-RNTI.
A. Preamble Transmission
D. User Separation and Contention Resolution Let R depict the total number of available preambles in
1) User separation based on SIC: The decoding order of a RA slot. Consider a specific preamble r and let Yri be a
UEs in a NORA group is consistent with the power back-off random variable which takes value 1 if the preamble is used
order, i.e. the UE with the strongest received power will be by exactly i out of m UEs and 0 otherwise. It readily follows
decoded first, which is reasonable due to the SIC assumption. that [4], m−i
It’s easily seen that the Power BO indication in RAR already m 1 1
E Yri =
implies the decoding order of UEs in a NORA group, thus 1 − . (5)
i Ri R
unlike [33], the base station doesn’t need an individual control
First, we consider the scenario where one preamble is used by
channel to inform the UEs of the assigned order. After
two UEs, i.e. i = 2.
decoding the UE’s message, the base station identifies the temp
C-RNTI and receives the initial layer 3 message. The UEs are assumed to be uniformly distributed in the
2) CRI transmission: The final stage of the RA procedure cell, thus the time interval between two UEs’ arrivals ∆t is
is contention resolution, which is based upon the base station depicted as
|d1 − d2 |
responding with a Contention Resolution Identity (CRI) mes- ∆t = , (6)
sage. Similar to the second stage the base station first transmits c
a Downlink Control Information (DCI) message to specify where c = 3×108m/s while dj (j = 1, 2) depicts the distance
the DL RBs for PDSCH transmission. The PDSCH includes between the base station and the j-th UE, which follows the
the Logical Channel Identity (LCID) Medium Access Control distribution given by [37]
(MAC) subheader and CRI MAC control element, as shown in 2x
Fig. 1. The LCID field is used to identify the subsequent CRI fdj (x) = , 0 < x < dc , (7)
dc 2
MAC control element while the CRI MAC control element
is utilized to reflect the initial layer 3 message sent by the where dc denotes the cell radius.
UE. The CRI contains configuration information regarding From Eq. (7), the distribution of ∆t can be readily computed
subsequent data transmission. and given by
The UE starts a CR timer (length determined by MAC 4c dc
f∆t (y) = 4 2d3c − 3d2c cy + c3 y 3 , 0 < y < .
Contention Resolution Timer broadcast on PBCH) after trans- (8)
3dc c
6
p(t)
U1[1]
U2[1] Uk[1]
t
TAP TW TRA
TRAI
Transmission interval
TRAP for k-th RA slot
Ă Ă t
0 t1 t2 tk-1 tk
RA slot data
d
Tbackoff
WBO
According to Section II. C, the UEs with successful pream- where p(t) is the arrival distribution
R T and tk is the start of the
ble transmission will receive the RAR message and transmit k-th RA slot. p(t) follows that 0 AP p(t)dt = 1.
the initial layer 3 message. In particular, the UEs in a NORA The UEs with preamble or message transmission failure
group will transmit their messages in the same resource blocks. will perform random backoff before returning to preamble
However, due to channel distortion, the decoding of the layer transmission. The number of contending UEs that transmit
3 message may not be successful. As a result, the UEs with their l-th (l ≥ 2) preamble in the k-th RA slot contains two
unsuccessful message transmission will return to preamble parts. The first part originates from the UEs whose (l − 1)-
transmission. th preamble transmission failed (i.e. Uk′ ,P F [l − 1]) in the
′
The UEs which have successfully transmitted their pream- k RA slot. Among these faild UEs, pk′ ,k of them end up
bles are divided into two parts, i.e. transmitting the l-th preamble in the k-th RA slot after the
1 2 random backoff process. Since these UEs perform uniform
Uk,P S [l] = Uk,P S [l] + Uk,P S [l] , (16)
backoff within the backoff window WBO (length determined
1 by BI in the RAR message), the value of pk′ ,k differs regarding
where Uk,P S [l] corresponds to UEs with no collision and ′ ′ ′ ′
2
Uk,P S [l] corresponds to UEs in NORA groups, which are k (kmin ≤ k ≤ kmax ). The second part originates from the
8
1 pout,1 + pout,2 2
Uk,MS [l] = (1 − pout,0 ) Uk,P
[l] + 1 −
S Uk,P S [l]
2
ps2 (U − 1) (19)
φ0 Uk α1 − φ12 φ2
k Uk
= e− 2θ2 Uk [l] pl e− R + e 2θ 1 + e− 2θ2 Uk [l] pl e− R .
2 2 (R − 1)
PRACH PRACH
time
TPF0 Tbackoff
TPRACH TPD WRAR WBO
(a) Preamble failure
Fig. 5. The delay analysis regarding (a) preamble failure, (b) preamble success but message failure and (c) preamble success and message success.
(tk′ +TP F 0 +WBO )−tk−1 +WBO ′
+WBO
WBO , (k − 1) − TP FT0RAP ≤ k ≤ k − TP FT0RAP
′
TRAP TP F 0 +WBO TP F 0
WBO , k− < k < (k − 1) − TRAP
pk′ ,k = TRAP (25)
tk −(tk′ +TP F 0 )
′
,
TP F 0
(k − 1) − TRAP ≤ k ≤ k − TTRAP PF0
WBO
ortherwise
0,
(tk′′ +TM F 0 +WBO )−tk−1 +WBO ′′
+WBO
WBO , (k − 1) − TM FT0RAP ≤ k ≤ k − TM FT0RAP
′′
TRAP TM F 0 +WBO TM F 0
WBO , k− < k < (k − 1) − TRAP
pk′′ ,k = TRAP (28)
tk −(tk′′ +TM F 0 )
′′
, (k − 1) − TRAP ≤ k ≤ k − TTM
TM F 0 F0
RAP
WBO
ortherwise
0,
K X
L
,K L
X XX 3 message as well as waiting time before the CR timer expires.
pMF = Uk,MF [l] Uk,P S [l]. (32)
k=1 l=1 k=1 l=1 TP F = TP F 0 + Tbackof f , (33)
TMF = TMF 0 + Tbackof f , (34)
where
TP F and TMF represents the average time required for TP F 0 = TP RACH + TP D + WRAR . (35)
a failed preamble transmission and a failed message trans-
mission with successful preamble transmission respectively, TMF 0 =TP RACH + TP D + ηWRAR + TRAR
. (36)
as depicted in Eq. (33) and Eq. (34). As illustrated in Fig. + TR−3 + TMsg3 + WCR
5, TP F contains the transmission time for PRACH preamble,
processing time for the base station to detect preamble, waiting As for TS , since the UE is capable of receiving the CR
time before the RA response window ends and the backoff message before the CR timer expires, the fraction effect ξ(0 <
time before preamble retransmission. Regarding TMF , with ξ < 1) is considered. TS contains the transmission time for
the successful preamble transmission, the UE is able to receive CR message, as shown in Eq. (37).
the RAR message before the WRAR timer ends. η(0 < η < 1) TS =TP RACH + TP D + ηWRAR + TRAR
is introduced to model this effect. Compared to TP F , TMF . (37)
+ TR−3 + TMsg3 + ξWCR + TCR
adds the transmission time for RAR and layer 3 message, the
time interval between a UE receiving a RAR and sending layer Since UEs perform random uniform backoff, Tbackof f fol-
10
35
of preambles minus the number of idle preambles and suc-
NORA
ORA
cessfully received preambles. Hence, PC is given as
Maximum throughput 20
PC = R − R E Yr |Uk − Uk,P S KR. (42)
20 achieved at m=53/54 k=1
15
4) Access Success Probability: PS is defined as the prob-
10 ability that a UE successfully completes the RA procedure
5
within the maximum number of preamble transmissions. That
is, PS is the ratio between total number of successfully
0
0 50 100 150 200 250 300 accessed UEs and the total number of UEs arrived in TP ,
Number of UEs which perform preamble transmission
expressed as
Fig. 6. Throughput of preamble transmission given ps2 = 0.6 and R = 54. PS = RRA /U . (43)
5) CDF of the Number of Preamble Transmissions: F (m)
lows a uniform distribution within [0, WBO ]. Thus, is defined as the CDF of the number of preamble transmissions
to perform a RA procedure for the successfully accessed UEs,
1
Tbackof f = WBO . (38) where m is the number of preamble transmissions. More
2 specifically, it is the ratio between the number of successfully
IV. P ERFORMANCE E VALUATION accessed UEs which transmit no more than m preambles and
the number of all successfully accessed UEs, given as
A. Performance Metrics
K X m
,K L
1) Throughput of Preamble Transmission: Consider a cer- X XX
F (m) = Uk,MS [l] Uk,MS [l]. (44)
tain RA slot in which a total number of m UEs transmit their
k=1 l=1 k=1 l=1
preambles. The throughput of preamble transmission RP is
defined as the number of successful transmitted preambles 6) CDF of the Access Delay: G(d) is defined as the CDF
given m simultaneous preamble transmissions. In the NORA of the delay between the first preamble attempt and the
scheme, it’s readily given from Eq. (11) as completion of the RA process for the successfully accessed
m−1 UEs. That is, G(d) is the ratio between the number of the
ps2 (m − 1)
N ORA 1 successfully accessed UEs whose access delay is no greater
RP = E [S] =m 1 + 1− .
2 (R − 1) R than d and the total number of successfully accessed UEs.
(39) G(d) is estimated by
In the ORA scheme, the throughput is expressed as
X(d)
K mmax
,K L
m−1
1
X XX
RPORA = m 1 − (40) G(d) = Uk,MS [l] Uk,MS [l], (45)
R k=1 l=1 k=1 l=1
which states that a UE succeeds in the preamble transmission where mmax (d)(mmax (d) ∈ N ) is the maximal number of
if all the other m − 1 UEs select the other R − 1 preambles. preamble transmissions by an UE and is estimated by setting
As illustrated in Fig. 6, the maximal throughput of preamble l = mmax (d) in Eq. (29) and let Tl = d.
transmission 20 is achieved when m equals to 53 or 54 in the 7) Average Number of Preamble Transmissions for Success-
ORA scheme. In comparison, more than 30% of throughput fully Accessed UEs: L is defined as the ratio between the
improvement can be achieved when m equals to 69 in the total number of preamble transmissions for all the successfully
NORA scheme. accessed UEs and the total number of all the successfully
2) Throughput of RA Process: RRA is defined as the total accessed UEs. Hence, L is given as
number of successfully accessed UEs in the RA time interval
K X m
,K L
TRAI , i.e. maximum supported UEs. That is, RRA is the X XX
L̄ = Uk,MS [l] · l Uk,MS [l]. (46)
accumulated number of UEs with successful preamble and
k=1 l=1 k=1 l=1
message transmission, expressed as
8) Average Access Delay for Successfully Accessed UEs:
K X
L
X DRA is defined as the ratio between the total access delay of
RRA = Uk,MS [l]. (41)
preamble transmissions for all the successfully accessed UEs
k=1 l=1
and the total number of all the successfully accessed UEs.
3) Collision Probability: PC is defined as the ratio between DRA is estimated by
the number of occurrences when two and more UEs make
X(d)
K mmax
,K L
a RA attempt using exactly the same preamble (resulting in X XX
undetected collision) and the overall number RA preambles DRA = Uk,MS [l] · Tl Uk,MS [l]. (47)
k=1 l=1 k=1 l=1
(with or without access attempts) in the period. That is,
PC is the ratio between the number of undetected collided Among the aforementioned performance metrics, the
preambles and the overall number of preambles. The number throughput of RA process RRA , collision probability PC ,
of undetected collided preambles is equal to the total number access success probability PS and average access delay DRA
11
10 5 20 400
R=4.0e+04
R=3.0e+04
R=2.0e+04
NORA ORA
Capacity of RA process R=1.5e+04
Uk,S
Uk,F
10 4 10 200
Uk,S
Uk,F
10
B. Simulation Analysis 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
0
k-th RA slot
10 0 1
0.9
0.6
10 -2 0.5
Ana. NORA, Traffic Model 1
Ana. ORA, Traffic Model 1 0.4 Ana. NORA, Traffic Model 1
Ana. NORA, Traffic Model 2 Ana. ORA, Traffic Model 1
Ana. ORA, Traffic Model 2 0.3 Ana. NORA, Traffic Model 2
Sim. NORA, Traffic Model 1 Ana. ORA, Traffic Model 2
10 -3
Sim. ORA, Traffic Model 1 Sim. NORA, Traffic Model 1
0.2
Sim. NORA, Traffic Model 2 Sim. ORA, Traffic Model 1
Sim. ORA, Traffic Model 2 Sim. NORA, Traffic Model 2
0.1
Sim. EAB T S =320, T P =800 [38] Sim. ORA, Traffic Model 2
10 -4 0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 1 2 3 4 5 6 7 8 9 10
Total number of arrived UEs ×10 4 Number of preamble transmission
1 1
0.9 0.9
0.8 0.8
Access success probability
0.7 0.7
0.5 0.5
Ana. NORA, Traffic Model 1
0.4 Ana. ORA, Traffic Model 1 0.4 Ana. NORA, Traffic Model 1
Ana. NORA, Traffic Model 2 Ana. ORA, Traffic Model 1
0.3 Ana. ORA, Traffic Model 2 0.3 Ana. NORA, Traffic Model 2
Sim. NORA, Traffic Model 1 Ana. ORA, Traffic Model 2
0.2 Sim. ORA, Traffic Model 1 0.2 Sim. NORA, Traffic Model 1
Sim. NORA, Traffic Model 2 Sim. ORA, Traffic Model 1
0.1 Sim. ORA, Traffic Model 2 0.1 Sim. NORA, Traffic Model 2
Sim. EAB T S =320, T P =800 [38] Sim. ORA, Traffic Model 2
0 0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 50 100 150 200 250 300 350 400 450 500
Total number of arrived UEs ×10 4 Access delay
Fig. 9. Collision and access success probability of the NORA and ORA Fig. 10. CDF of the number of preamble transmissions and access delay
schemes under both Traffic Models. for the successfully accessed UEs in NORA and ORA schemes under both
Traffic Models.
TABLE II
A CCESS DELAY COMPARISON
U
5000 10000 15000 20000 25000 30000 35000 40000 45000 50000
DRA (ms)
NORA 38 40 45 55 68 69 69 69 68 67
ORA 39 43 54 71 71 71 70 69 68 69
EAB 31 3985 4654 4359 3949 3604 3023 2502 2166 1908
6
access scenarios is one of our main research directions.
Ana. NORA, Traffic Model 1
Average number of preamble transmission
4
Sim. ORA, Traffic Model 2 In this paper, we have proposed the NORA scheme to
L=3.4, U=20000
3.5
alleviate the potential access congestion problem regarding the
3 L=2.5, U=20000
massive-connection scenarios in 5G networks. Specifically, the
2.5
spatial distribution characteristics of UEs were utilized to real-
2 L=2.5, U=35000
ize multi-preamble detection and RAR reception, which effec-
1.5
tively improves the preamble transmission success probability.
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Total number of arrived UEs ×10 4
Moreover, NORA allows simultaneous message transmission
(a) Average L
of multiple UEs, thus alleviates the demand on limited PUSCH
resources. In addition, we have presented the analytical model
120
Ana. NORA, Traffic Model 1
to investigate the transient behavior of the NORA process
110 Ana. ORA, Traffic Model 1
Ana. NORA, Traffic Model 2
with non-stationary arrivals under realistic assumptions. Be-
100 Ana. ORA, Traffic Model 2
Sim. NORA, Traffic Model 1 sides, a comprehensive evaluation of our proposition is given,
Average access delay
denotes the wavelength. vj denotes the Rayleigh channel gain [6] Evolved Universal Terrestrial Radio Access (E-UTRA); Physical
with the probability density function (PDF) of |vi | given by Channels and Modulation, 3GPP TS 36.211 V10.4.0, Dec.
2011.
x x2
[7] S. Stefania, T. Issam, and B. Matthew, LTE-The UMTS Long
f|vj | (x) = 2 e− 2θ2 , 0 ≤ x < ∞, (50)
θ Term Evolution: From Theory to Practice. Wiley, 2011.
[8] A. Laya, L. Alonso, and J. Alonso Zarate, “Is the random access
where θ2 represents the variance. channel of LTE and LTE-A suitable for M2M communications?
At the base station, the messages of UEs within a NORA A survey of alternatives,” IEEE Commun. Surveys Tuts., vol. 16,
group are decoded in a successive way based on the decoding no. 1, pp. 4–16, Oct. 2014.
order, i.e. power back-off order. Thus, the achievable data rate [9] Z. Wang and V. W. Wong, “Optimal access class barring for
of the two UEs in a NORA group is given by Eq. (51) and stationary machine type communication devices with timing
advance information,” IEEE Trans. Wireless Commun., vol. 14,
Eq. (52) respectively. no. 10, pp. 5374–5387, Oct. 2015.
[10] D. T. Wiriaatmadja and K. W. Choi, “Hybrid random access and
!
2
P1 |h1 | data transmission protocol for machine-to-machine communi-
R1 = log 1 + 2 , (51)
P2 |h2 | + σ 2 cations in cellular networks,” IEEE Trans. Wireless Commun.,
! vol. 14, no. 1, pp. 33–46, Jan. 2015.
2 [11] S. Duan, V. Shah-Mansouri, Z. Wang, and V. Wong, “D-ACB:
P2 |h2 |
R2 = log 1 + , (52) Adaptive congestion control algorithm for bursty M2M traffic
σ2 in LTE networks,” IEEE Trans. Veh. Technol, vol. 65, no. 12,
pp. 9847–9861, Dec. 2016.
where σ 2 denotes the variance of noise. [12] F. Morvari and A. Ghasemi, “Two-Stage resource allocation for
Let R̂j denote the target data rate of the j-th UE. The random access M2M communications in LTE network,” IEEE
o the j-th UE’s message can be defined
successful ndetection of Commun. Lett., vol. 20, no. 5, pp. 982–985, May 2016.
as Zj = Rj ≥ R̂j . Given that the detection of the j-th [13] J.-B. Seo and V. C. Leung, “Design and analysis of backoff
algorithms for random access channels in UMTS-LTE and IEEE
UE’s message is based on the successful decoding of the prior 802.16 systems,” IEEE Trans. Veh. Technol, vol. 60, no. 8, pp.
(j − 1) UEs’ messages, the outage probability of the j-th UE 3975–3989, Oct. 2011.
is expressed as [33] [14] T.-M. Lin, C.-H. Lee, J.-P. Cheng, and W.-T. Chen, “PRADA:
Prioritized random access with dynamic access barring for MTC
pout,j = 1 − P (Z1 ∩ · · · ∩ Zj ) , (53) in 3GPP LTE-A networks,” IEEE Trans. Veh. Technol, vol. 63,
no. 5, pp. 2467–2472, Jun. 2014.
Following the similar derivations in [33], the outage prob- [15] Y. C. Pang, S. L. Chao, G. Y. Lin, and H. Y. Wei, “Network
abilities of the 1-st and 2-nd decoded UEs in a NORA group access for M2M/H2H hybrid systems: A game theoretic ap-
are given as proach,” IEEE Commun. Lett., vol. 18, no. 5, pp. 845–848, Jun.
2014.
φ1
pout,1 = 1 − α1 e− 2θ2 , (54) [16] J. Chen, Y. T. Lin, and R. G. Cheng, “A delayed random
φ +φ
access speed-up scheme for group paging in machine-type
− 12θ2 2 communications,” in 2015 IEEE International Conference on
pout,2 = 1 − α1 e , (55)
Communications (ICC), Jun. 2015, pp. 623–627.
where α1 = 2
and φj = σ2
2R̂j − 1 for [17] Study on RAN Improvements for Machine-type Communica-
δ
1+10− 10 (2R̂1 −1) Pj l2j tions, 3GPP TR37.868 V11.0.0, Sep. 2011.
j = 1, 2. [18] H. Thomsen, N. K. Pratas, Č. Stefanović, and P. Popovski,
Regarding the UEs with no collision, the outage probability “Code expanded radio access protocol for machine-to-machine
communications,” Trans. Emerging Tel. Tech., vol. 24, no. 4,
is given as pp. 355–365, Jun. 2013.
[19] H. S. Jang, S. M. Kim, K. S. Ko, J. Cha, and D. K. Sung,
( ! )
2
P0 |h0 | φ0
pout,0 = P log 1 + 2
≥ R̂0 = 1 − e− 2θ2 , (56) “Spatial group based random access for M2M communications,”
σ IEEE Commun. Lett., vol. 18, no. 6, pp. 961–964, Jun. 2014.
[20] K. S. Ko, M. J. Kim, K. Y. Bae, D. K. Sung, J. H. Kim, and
2
where φ0 = Pσ0 l2 2R̂0 − 1 . J. Y. Ahn, “A novel random access for fixed-location machine-
0 to-machine communications in OFDMA based systems,” IEEE
Commun. Lett., vol. 16, no. 9, pp. 1428–1431, Sep. 2012.
R EFERENCES [21] D. Niyato, P. Wang, and D. I. Kim, “Performance modeling
[1] Y. Liang, X. Li, J. Zhang, and Y. Liu, “A novel random and analysis of heterogeneous machine type communications,”
access scheme based on successive interference cancellation IEEE Trans. Wireless Commun., vol. 13, no. 5, pp. 2836–2849,
for 5G networks,” in 2017 IEEE Wireless Communications and May 2014.
Networking Conference (WCNC), Mar. 2017, pp. 1–6. [22] T. Kim, H. S. Jang, and D. K. Sung, “An enhanced random
[2] J. G. Andrews, S. Buzzi, W. Choi, S. V. Hanly, A. Lozano, access scheme with spatial group based reusable preamble
A. C. Soong, and J. C. Zhang, “What will 5G be?” IEEE J. allocation in cellular M2M networks,” IEEE Commun. Lett.,
Sel. Areas Commun., vol. 32, no. 6, pp. 1065–1082, Jun. 2014. vol. 19, no. 10, pp. 1714–1717, Oct. 2015.
[3] Evolved Universal Terrestrial Radio Access (E-UTRA); Overall [23] Y. Yu and G. B. Giannakis, “High-throughput random access
Description; Stage 2, 3GPP TS 36.300 V13.2.0, Jan. 2016. using successive interference cancellation in a tree algorithm,”
[4] C. H. Wei, G. Bianchi, and R. G. Cheng, “Modeling and IEEE Trans. Inf. Theory, vol. 53, no. 12, pp. 4628–4639, Dec.
analysis of random access channels with bursty arrivals in 2007.
OFDMA wireless networks,” IEEE Trans. Wireless Commun., [24] P. Salvo Rossi, K. Kansanen, R. R. Muller, and C. Rachinger,
vol. 14, no. 4, pp. 1940–1953, Apr. 2015. “Power randomization for iterative detection over random-
[5] Evolved Universal Terrestrial Radio Access (E-UTRA); Medium access fading channels,” IEEE Trans. Wireless Commun.,
Access Control (MAC), 3GPP TS 36.321 V9.3.0, Jun. 2010. vol. 14, no. 10, pp. 5704–5713, Jun. 2015.
15