A survey on NB-IoT downlink scheduling: Issues and potential solutions

Full text

A Survey on NB-IoT Downlink Scheduling: Issues
and Potential Solutions
Rubbens Boisguene∗, Sheng-Chia Tseng†, Chih-Wei Huang∗, and Phone Lin‡
∗Department of Communication Engineering, National Central University, Taoyuan, Taiwan
Email: rubbensonly@gmail.com, cwhuang@ce.ncu.edu.tw
†Smart Network System Institute, Institute for Information Industry, Taipei, Taiwan
Email: joetseng@iii.org.tw
‡Department of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan
Email: plin@csie.ntu.edu.tw
Abstract—The NarrowBand Internet of Things (NB-IoT) is one
of the most promising technologies that fits the requirements of
the low-power wide area networks (LPWAN). NB-IoT targets
on supporting communications for small amount of data during
a relatively long period of time (i.e., delay tolerant), which is
one of the key features of IoT applications. Several studies have
been conducted by some standard working groups to define
the standards for NB-IoT. For example, the working group
3GPP aims to provide communication protocols with low energy
consumption, good coverage penetration, and so on. Compared
with the existing wireless communication protocols, less spectrum
is allocated for NB-IoT. Thus, how to more efficiently use the
resource/spectrum (i.e., resource allocation and scheduling) is one
of the key issues. In this paper, we first describe the design of
the physical layer for the downlink communication, especially the
scheduling process. Secondly, we discuss the issues for resource
allocation in NB-IoT while the delay requirement in the physical
layer is satisfied. Thirdly, we discuss the open questions and
possible solutions.
Index Terms—NarrowBand Internet of Things (NB-IoT); re-
source allocation; scheduling
I. INTRODUCTION
The NarrowBand Internet of Things (NB-IoT) is one of the
most promising technologies that fits the requirements of the
low-power wide area networks (LPWAN). It has drawn many
researchers and working groups attention. The working group
3GPP has started to work on the specifications for NB-IoT
since 2014. Initially, the specification of NB-IoT focuses on
low-end, low-power devices called Cat NB1. The specification
of NB-IoT also allows devices to send small amount of data
in parallel [1]. From the technical point of view, NB-IoT is
designed for low-cost devices, operated on a low-frequency
bandwidth of 180 kHz to offer a network coverage over 160
dB with a latency around 10 seconds [2]. With these features,
NB-IoT can target IoT devices that are delay tolerant or located
in areas where signal transmission is bad.
With the massive deployment of various batteries operated
IoT devices in those remote areas, NB-IoT is more suited to
reach them. NB-IoT is therefore considered as a better option
when it comes to LPWAN technologies. The deployment of
NB-IoT on these areas supports a number of IoT services such
as smart agriculture, smart city, smart home, smart metering
and so on [3]. Further enhancements will be proposed in the
release 14 of the 3GPP specifications.
Although NB-IoT has these advantages, there are still issues
that need to be resolved. For example, optimization works have
been proposed to deal with eventual data load and congestion
problems [4]. In this paper, we provide a close look at the
downlink design for NB-IoT. In addition, we highlight the
scheduling issues related to resource allocation for UEs in the
downlink phase given standardized parameter selection.
The rest of this paper is organized as follows. Section II
introduces the NB-IoT and its features. Section II-B provides
general details of the NB-IoT MAC and physical layers.
The NB-IoT scheduling process is presented in Section III.
The issue related to the scheduling process is highlighted in
Section IV. The potential solutions presented in Section V.
We conclude the paper in Section VI.
II. NB-IOT DOWNLINK FEATURE OVERVIE W
By reusing the existing network structures (e.g., LTE or
GSM), NB-IoT provides greater flexibility for the deployment,
which makes it well suited for the 5G network [5]. The
requirements of NB-IoT are elaborated as follows:
•Low power consumption: Both power saving mode (PSM)
and extended discontinuous reception (eDRX) are ap-
plicable for NB-IoT, which provides a maximization of
battery life reaching 10 years [6].
•Low channel bandwidth: The channel bandwidth is 200
kHz (180 kHz plus guard bands), which also makes it
suitable for GSM channel re-farming because it allows
one GSM/GPRS channel to be replaced with a single NB-
IOT channel. NB-IoT operates on licensed bands with a
great flexibility [5].
•Low deployment cost: The capability of NB-IoT to be
integrated into the LTE network or to reuses the existing
GSM bands makes it easy for Mobile Network Operators
(MNOs) to deploy it. The complexity reduction helps a
lot so that it can run on a narrowband network. NB-IoT
can coexist with traditional cellular networks while being
simpler and serving more devices.
•Low cost for UE: The NB-IoT operates on 180 kHz.
The NB-IoT UE is also called CAT-NB1. The front-end
978-1-5090-4372-9/17/$31.00 ©2017 IEEE 547

TABLE I: NB-IoT Configuration
Uplink peak rate 250 kbps
Downlink peak rate 227 kbps
Bandwidth 180 kHz
UpLink SC-FDMA (15 KHz and 3,75 KHz)
Downlink OFDMA (15 KHz) single or double antenna
Duplex model Half duplex
MCL (Coverage) 164 dB (<15Km)
Transmit power 23 dBm
Power saving PSM, ext-IDRX, C-DRX
and digitizer of NB-IoT operate on 200 kHz and are
much simpler than that of LTE operating on 1.4 MHz.
Obviously, the low complexity leads to low device unit
cost, the CAT-NB1 UE is considered as cheaper than the
existing UE (around USD 5) [7].
•Support delivery of IP and non-IP data: The NB-IoT
supports both IP data delivery and non-IP data delivery
(NIDD). The NIDD is enhanced in several aspects, as the
SMS service may also be used to deliver data without the
use of the IP protocol.
•Extended coverage: With 164 dB MCL, NB-IoT provides
20dB additional link budget, enabling about ten times of
the better area coverage (e.g., 20 dB better compared to
GPRS) [8].
•Support for a massive number of devices: With its low
bandwidth and its extended coverage feature, the NB-
IoT technology is surely designed to meet the connec-
tivity requirements for massive MTC applications and
devices [9].
A. Operation Mode
The NB-IOT delivers an effective narrowband operation
with 180 kHz bandwidth for both the uplink and downlink,
which corresponds to one resource block in LTE network
configuration. The NB-IoT can be operated in 3 different
modes as shown in Fig. 1:
•In-band operation: It exists within the LTE carrier, which
one resource block in the LTE network is reserved for
NB-IoT.
•Guard band operation: It operates in the guard band
immediately adjacent to the LTE carrier, without affecting
the capacity of the LTE carrier. With this variety of
choices, the operator can choose the most suitable opera-
tion mode to satisfy its network performance requirement
while offering the services to IoT applications [10].
•Stand-alone operation: It uses re-farmed GSM low band
already existing in many countries (700MHz, 800MHz,
and 900MHz).
B. The Design of the Downlink for the MAC and PHY Layer
This section describes the design of the downlink of the
MAC and PHY layers of NB-IoT and identifies the related
scheduling issue that may occur.
The NB-IoT MAC and PHY downlink designs are similar
to that of the LTE. The NB-IoT downlink structure is designed
Fig. 1: NB-IoT Operation mode.
Fig. 2: NB-IoT Radio frame structure.
TABLE II: subcarrier spacing
Spacing 15 kHz
Subcarriers 12
1 frame 10 subframes
1 subframe 14 OFDM symbols
FTT length 128 samples = 66.7 sec
Sampling rate 1.92 MHz (15 kHz * 128)
with 15 kHz of subcarrier spacing (see Table I). The down-
link waveform is based the OFDMA modulation. The frame
structure in NB-IoT is briefly described as shown in Fig. 2.
The NB-IoT is based on the Orthogonal Frequency-Division
Multiple Access (OFDMA) with the frame structure of 180
kHz bandwidth. A frame consists of 10 subframes with length
1 ms, i.e., the transmission time interval (TTI) is 1 ms. The
lengths of a timeslot, subframe, and frame are 0.5 ms, 1 ms,
and 10 ms, respectively, which are identical to LTE [11].
The NB-IoT supports the operation with either one or two
antenna ports, AP0 and AP1, the same transmission scheme
applied to NPBCH, NPDCCH, and NPDSCH [12]. An NB-
IoT carrier uses one LTE Physical Resource Block (PRB) in
the frequency domain, i.e., 12 subcarriers of 15 kHz for a total
bandwidth of 180 kHz as shown in Table II.
1) NB-IoT Downlink Channels: The NB-IoT downlink
structure has almost the same channels as that in LTE. The
definition of the channels are given below:
a) NPBCH: The NPBCH stands for Narrowband Phys-
ical Broadcast Channel. The NPBCH carries the critical in-
formation and parameters used for the initial access of the
cell. The information and parameters are carried by the Master
Information Block (MIB). NB-IoT is based on the quadrature
phase shift keying (QPSK) modulation scheme.
b) NPDCCH: The Narrowband Physical Downlink Con-
trol Channel (NPDCCH) indicates the IDs of the UEs having
data transmitted in the NPDSCH. It carries some control infor-
mation such as paging, UL/DL assignment, RACH response,
548

Fig. 3: Frame Diagram
ACK to push, power control, and so on. It controls the data
transmission between the eNB and the UEs. It carries the
scheduling assignments and other control information in the
form of DCI messages. The interval between start of two
NPDCCH is PDCCH period (pp).
c) NPDSCH: The Narrowband Physical Downlink
Shared Channel (NPDSCH) is the main data bearing channel.
It carries UE unicast data, some control information and the
broadcasted system information, i.e., SIBs.
2) NB-IoT Downlink Signals: Two types of signals are
defined in the NB-IOT configuration: the Narrowband Ref-
erence Signal (NRS) and the synchronization signals. NRS
is transmitted through all broadcast and downlink subframes.
The synchronization signals are generated for cell search.
A UE acquires time and frequency synchronization with
a cell and detects the physical layer ID through cell search
procedures. Two types of synchronization signals are the
primary synchronization signal (NPSS) and the secondary
synchronization signal (NB-SSS):
a) NB-PSS: It carries the physical cell ID responsible for
cell synchronization. The signal is transmitted in the subframe
#5 using the last 11 OFDM symbols every 10 ms.
b) NB-SSS: It carries the physical layer cell identity
group radio frame synchronization. The signal is transmitted
in subframe #9 using the last 11 OFDM symbols every 20 ms.
C. Configuration of Channels and Signals
The downlink channels and signals are configured to es-
tablish frame structure and time settings. Data and signals
are transmitted through channels containing particular system
information and time control for the next operation between
the eNB and specific IoT devices [13]. The components are
presented as follows (see Fig. 3):
•Narrowband Master Information Block (MIB-NB): The
MIB-NB is transmitted over the NPBCH. Located in
the first subframe indexed subframe #0, the 34-bit MIB-
NB contains network operation modes. Using subframe
#0 can also avoid collision between the NPBCH and
potential MBSFN transmission under in-band operation.
The contained information includes system timing, access
barring, and scheduling details for the upcoming system
information blocks.
•Narrowband System Information Block (SIB-NB) A min-
imum of 2 SIB-NBs is required for a UE to access
the eNB, i.e., SIB1-NB and SIB2-NB. Transmitted over
the NPDSCH, SIB-NBs carry relevant information for
a UE to access a cell or perform cell re-selection. Cell
status and reservations are indicated in SIB1-NB using
subframe #4 in each radio frame [14]. SIB2-NB contains
the radio resource configuration information.
•UE Search spaces The UE search spaces are divided
into two groups, the common search space, and the UE-
specific search space, and required to be monitored. There
can be overlap between two groups of search spaces for
a UE. The UE-common search space itself is divided
into 2 types: type-1 is used for paging, while type-2 is
used for random access. The UE-specific search space
can carry Downlink Control Information (DCIs) for UE-
specific allocations, while the common search space can
carry DCIs common for all UEs.
III. THE SCHEDULING PRO CE SS
The scheduling process allocates radio resources in PRB
to UEs every transmission time interval (TTI). The NB-
IoT scheduling process is located in lower MAC and upper
PHY layer. One of the key elements of the network is the
ability to control and prioritize bandwidth across UEs. The
MAC scheduler is responsible for deciding how and when the
channels are used by the eNB and the UEs of a cell. During
the NB-IoT scheduling process, an offset is given to each UE
to avoid allocation overlapping.
NPDCCH is considered as the core element of the downlink
control channels as it carries DCI. The DCI contains uplink
scheduling grants, downlink scheduling assignments, and the
type of modulation being used for NPDSCH. DCI Format
N0 is used for the uplink grant. Assignments of PDSCH is
transmitted through DCI format N1. It is used for all NPDSCH
(user data and NB-SIBs) except NPDSCH carrying paging.
DCI format N2 is used for paging and direct indications [15].
NPDSCH carries UE’s downlink data and control information.
When receiving NPDCCH, the UE searches the DCI in-
tended for it according to the UE search spaces. The DCI
contains the scheduling delay between the end of NPDCCH
transmission and the beginning of NPDSCH or NPUSCH for a
UE. A different offset index represented by k0is given to each
UE to avoid overlapping. The offsets are selected from the
offset pool represented in Table III. Assigning k0is a nontrivial
task, as the selection of offsets is very limited comparing to
the number of UEs to schedule.
IV. THE NB-IOT S CHEDULING ISS UE
The scheduler in the base station controls the allocation
of shared time-frequency resources among users at each time
549

TABLE III: k0for DCI format N1
Offset index k0
Rmax <128 Rmax ≥128
0 0 0
1 4 16
2 8 32
3 12 64
4 16 128
5 32 256
6 64 512
7 128 1024
Fig. 4: k0assignment for UEs.
instant. The reception of NPDSCH transmission in time-
frequency combination for each TTI is determined. The
scheduling decision provided by the scheduler must be com-
pliant with more limited NB-IoT specifications. The resource
assignment must be in a specific format taking into account
reserved signaling resources and capabilities of the UE.
The relevance of the resource allocation issue is further
explained by numerical details and presented as follow. Con-
sidering a scenario of smart metering communications in
which 100 smart meters that need to connect to the NB-IoT
network. The network is configured with the FFT length equal
to 128. Thus the resource allocation is made withing these 128
subframes started from subframe 0 to subframe 127. During
the NPDDCH period, the resource assignment process is made
via the DCI responsible for resource allocation. Given a smart
meter located indoor, the UE normally transmits at minimum
16 Bytes of data which is equal to 128 bits [16] [17]. NB-
IoT is QPSK modulated to facilitate low complexity decoding.
Following the details presented in the release 13 of the 3GPPP
specifications, 128 bits occupy 6 subframes [15]. With the
coverage enhancement mechanism, several repetitions occur
for enhanced decoding probability, i.e., the same data for each
UE is transmitted several times. Considering the need of 4
repetitions for the UE due to its location in an area of poor
signal strength, the number of occupied subframes increases by
4 times reaching a total of 24 subframes for the UE as shown
in Fig. 4. Given the chosen offset index 5 with k0equal to
32, the next smart meters is to be scheduled after 32 slots.
The total number of subcarriers for one resource block is 12
with a frequency spacing of 15 kHz (12*15 kHz = 180 kHz).
With a large number of UEs and a fixed number of available
subcarriers to be allocated, the resources are not guaranteed
even with successful random access procedures.
To find the proper scheduling time and the number of
accessible UEs, we start with calculation presented in [15]:
tn=n+5+k0(1)
SFN P D SC H =pp −SFN P D CC H (2)
where tndenotes the exact scheduled time for UE n, 5 is the
necessary delay for DL, and k0the offset value. In (2), pp
represents the overall number of subframes, SFN P D SC H and
SFN P D CC H denote the total number of subframes occupied
by the NPDSCH and the NPDCCH respectively.
Given 8 DCIs of 4 subframes each, we consider the NPD-
CCH use a total of 32 subframes. The NPDSCH follows
right after with each UE transferring data after the scheduling
calculation of (1). Following (2), 32 is deducted from the
total number of subframes which is 128 subframes. Thus the
number of available subframes for NPDSCH is 96. Without the
waste, 4 UEs can share the available resources as 96/24 = 4.
In this example, only 3 UEs are successfully scheduled as the
total number of wasted resources occupies 24 subframes. Thus
25% of the overall resource is wasted as observed in Fig. 4.
The situation mentioned above clearly explains how critical
the issue is. With the best possible k0and a successful
random access process, there is still a possibility of resource
shortage, because of the signaling overhead and the limited
selection of offsets under data repetitions. The configuration
and k0selection may lead to inefficient resource allocation
for the UEs without a proper scheduling scheme. The issue of
scheduling is presented as the need for optimization regarding
resource allocation.
Several works can be found in the literature with related
downlink scheduling mechanisms for LTE and UMTS. In [18],
a time-division-based cyclic scheduling method was proposed
for UMTS downlink shared channels. An elastic shared-
channel assignment and scheduling scheme to cope with
the UMTS scheduling-related problems were applied. Kihl
et al. [19] analyzed the performance of different downlink
scheduling strategies under various urban and rural scenar-
ios such as vehicle to infrastructure for safety data traffic
applications. The evaluation is based on the coexistence of
various scenarios and how the scheduling strategies to avoid
any collision on an LTE-based network. A comparison of
LTE downlink scheduling methods was presented in [20].
The authors studied the performance analysis of various LTE
downlink schedulers both in time and frequency domain.
Studies as mentioned above were mainly designed for
general LTE and UMTS, and the effectiveness on the NB-
IoT with massive MTC access is still an open issue. It is
important to take into account various NB-IoT parameters and
restrictions while conceiving the scheduling mechanism for an
optimal resource assignment.
V. POTENTIAL SOLUTIONS
Following the issue presented in Section IV, we raise open
design issues toward effective NB-IoT scheduling mecha-
nisms. An ideal NB-IoT scheduling process needs to consider:
550

TABLE IV: Number of scheduled UE
Number of repetitions 0 1 2 3
Straightforward (pp=128) 8 4 3 2
Straightforward (pp=256) 8 6 4 3
Pre-divided resource allocation (pp=128) 8 6 4 3
Pre-divided resource allocation (pp=256) 8 8 8 7
TABLE V: Percentage of resources utilization
Number of repetitions 1 2 3
Straightforward (pp=128) 66.6% 75.0% 66.6%
Pre-divided resource allocation (pp=128) 100% 100% 100%
Straightforward (pp=256) 42.8% 42.8% 42.8%
Pre-divided resource allocation (p=p256) 57.1% 85.7% 100%
•Offset index selection: So with the limited k0and varying
size of payloads, it is critical to adapt the scheduling
process for high resource utilization to accommodate
more devices at the same time.
•UE-specific and common search space configuration:
The configuration decides the timing of NPDCCH and
NPDSCH for different UEs. An optimized UE-specific
and common search space configuration can eventually
improve the overall scheduling efficiency.
In regard to the two considerations mentioned above, the pro-
posed resource allocation scheme uses a pre-divided resource
allocation strategy to accommodate the UEs. For each UE,
the offset index selection k0is therefore made to match the
pre-divided area considered as the resource allocation unit
following the formula expressed in (1). The process contains
various features such as the dynamic data slicing which splits
the requested resource (for data transmission ) of a UE to fit the
scattered available resources. This feature aims to maximize
the network’s resource utilization.
A. Simualation results
During in simple simulation with a total of 8 UEs to be
scheduled and a resource request of 8 subframes per UE for
each transmission, we observe under various situations (in-
creasing number of repetitions), the number of scheduled UE
is decreasing. It is also shown in table IV, the pre-divided re-
source allocation mechanism outperforms the straightforward
method. Another feature called ”cross pp” will be investigated
in our future work to maximize the number of scheduled UE.
In table V, the proposed method offers a maximization of
the resource utilization up to 100% under various circum-
stances because of the dynamic data slicing feature.
VI. CONCLUSION
In this paper, we present a detailed study of the NB-IoT
downlink scheduling structure. The scheduling process is first
explained with details of the physical channels and signals.
As one of the first downlink scheduling study specifically for
NB-IoT, the NB-IoT specific scheduling problems were raised
with key design direction provided. Existing LTE schedulers
need to be carefully revised due to limited resources and sim-
plified options. A pre-divided resource allocation mechanism
is briefly presented to deal with the scheduling problem for
NB-IoT.
ACKNOWLEDGMENT
The research is based on work supported by the Ministry
of Science and Technology (MOST) of Taiwan, under grant
number 104-2923-E-002-005-MY3.
REFERENCES
[1] Ericsson, “NB-IoT: a sustainable technology for connecting billions of
devices,” Ericsson, Tech. Rep., April 2016.
[2] R. Ratasuk, B. Vejlgaard, N. Mangalvedhe, and A. Ghosh, “NB-IoT sys-
tem for M2M communication,” in 2016 IEEE Wireless Communications
and Networking Conference, April 2016, pp. 1–5.
[3] Huawei, “NB-IoT - Enabling New Business Opportunities,” Huawei,
Tech. Rep.
[4] 3GPP TR 45.820, “Cellular system support for ultra-low complexity and
low throughput Internet of Things (CIoT),” TR.
[5] C. Hoymann, D. Astely, M. Stattin, G. Wikstrom, J. F. Cheng,
A. Hoglund, M. Frenne, R. Blasco, J. Huschke, and F. Gunnarsson,
“LTE release 14 outlook,” IEEE Communications Magazine, vol. 54,
no. 6, pp. 44–49, June 2016.
[6] J. P. Bardyn, T. Melly, O. Seller, and N. Sornin, “IoT: The era of LPWAN
is starting now,” in ESSCIRC Conference 2016: 42nd European Solid-
State Circuits Conference, Sept 2016, pp. 25–30.
[7] 3GPP TS 36.101, “LTE : Evolved Universal Terrestrial Radio Access
(E-UTRA); User Equipment (UE) radio transmission and reception ,”
3rd Generation Partnership Project , TS version 13.4.0 Release 13.
[8] Alcatel Lucent / Nokia, “LTE evolution for IoT connectivity,” Tech. Rep.
[9] Ericsson, “Cellular networks for massive IoT,” Ericsson, Tech. Rep., Jan
2016.
[10] A. Rico-Alvarino, M. Vajapeyam, H. Xu, X. Wang, Y. Blankenship,
J. Bergman, T. Tirronen, and E. Yavuz, “An overview of 3GPP enhance-
ments on machine to machine communications,” IEEE Communications
Magazine, vol. 54, no. 6, pp. 14–21, June 2016.
[11] X. Lin, A. Adhikary, and Y. P. E. Wang, “Random Access Preamble
Design and Detection for 3GPP Narrowband IoT Systems,” IEEE
Wireless Communications Letters, vol. 5, no. 6, pp. 640–643, Dec 2016.
[12] Rohde-schwarz., “Narrowband Internet of Things Whitepaper,” Rohde-
schwarz., Tech. Rep.
[13] 3GPP TR 36.802, “Evolved Universal Terrestrial Radio Access (E-
UTRA); NB-IOT; Technical Report for BS and UE radio transmission
and reception,” 3rd Generation Partnership Project (3GPP), TR 36.802.
[14] 3GPP TS 36.304, “Technical Specification Group Radio Access Net-
work; Evolved Universal Terrestrial Radio Access (E-UTRA); UE
procedures in idle mode,” TS, June 2016.
[15] 3GPP TS 36.213, “LTE; Evolved Universal Terrestrial Radio Access
(E-UTRA); Physical layer procedures,” TS 36.213, May 2016.
[16] R. P. Lewis and P. Igict, “Assessment of communication methods
for smart electricity metering in the U.K.” in 2009 IEEE PES/IAS
Conference on Sustainable Alternative Energy (SAE), Sept, pp. 1–4.
[17] E. Yaacoub and A. Kadri, “LTE radio resource management for real-time
smart meter reading in the smart grid,” in IEEE International Conference
on Communication Workshop (ICCW), June 2015, pp. 2000–2005.
[18] C. H. Gan, N. C. Perng, P. Lin, and T. W. Kuo, “Time-Division-Based
Cyclic Scheduling for UMTS High-Speed Downlink Shared-Channels,”
IEEE Transactions on Vehicular Technology, vol. 56, no. 4, pp. 2086–
2094, July 2007.
[19] M. Kihl, K. Br, P. Mahanta, and E. Coelingh, “3GPP LTE downlink
scheduling strategies in vehicle-to-infrastructure communications for
traffic safety applications,” in 2012 IEEE Symposium on Computers and
Communications (ISCC), July 2012, pp. 000 448–000 453.
[20] M. Carpin, A. Zanella, J. Rasool, K. Mahmood, O. Grndalen, and O. N.
sterb, “A performance comparison of lte downlink scheduling algorithms
in time and frequency domains,” in IEEE International Conference on
Communications (ICC), June 2015, pp. 3173–3179.
551