1
Enabling Grant-Free URLLC: An Overview of
Principle and Enhancements by Massive MIMO
Jie Ding, Mahyar Nemati, Shiva Raj Pokhrel, Ok-Sun Park, Jinho Choi, Senior Member, IEEE , and Fumiyuki
Adachi, Life Fellow, IEEE
Abstract—Enabling ultra-reliable low-latency communication
(URLLC) with stringent requirements for transmitting data
packets (e.g., 99.999% reliability and 1 millisecond latency)
presents considerable uplink transmission challenges. For each
packet transmission over dynamically allocated network radio
resources, the conventional random access protocols are based
on a request-grant scheme. This induces excessive latency and
necessitates reliable control signalling, resulting in overhead. To
address these problems, grant-free (GF) solutions are proposed
in the fifth-generation (5G) new radio (NR). In this paper, an
overview and vision of the state-of-the-art in enabling GF URLLC
are presented. In particular, we first provide a comprehensive
review of NR specifications and techniques for URLLC, discuss
underlying principles, and highlight impeding issues of enabling
GF URLLC. Furthermore, we briefly explain two key phenomena
of massive multiple-input multiple-output (mMIMO) (i.e., chan-
nel hardening and favorable propagation) and build several deep
insights into how celebrated mMIMO features can be exploited to
address the issues and enhance the performance of GF URLLC.
Moving further ahead, we examine the potential of cell-free (CF)
mMIMO and analyze its distinctive features and benefits over
mMIMO to resolve GF URLLC issues. Finally, we identify future
research directions and challenges in enabling GF URLLC with
CF mMIMO.
Index Terms—URLLC, grant-free random access, retransmis-
sion, massive MIMO, cell-free.
I. INTRODUCTION
The excitement about fifth-generation (5G) is ushering in
the possibility of dramatic network improvements and moti-
vating service providers to plan for the future [1]. Today’s
Internet of Things (IoT) and networks support a broad range
of services by providing ubiquitous all-purpose connectives
[2]. 5G networks are expected to support wireless connections
originated by various IoT devices in addition to traditional
broadband services. To this end, three new service categories
were defined by the 3rd generation partnership project (3GPP)
This work was supported by the Institute for Information and Commu-
nications Technology Promotion (IITP) funded by the Korea Government
through the Ministry of Science, ICT and Future Planning (MSIT) through
the Development of Beyond 5G Mobile Communication Technologies (Ultra-
Reliable, Low-Latency, and Massive Connectivity) and Combined Access
Technologies for Cellular Based Industrial Automation Systems under Grant
2017-0-00724.
Jie Ding (Corresponding author), Mahyar Nemati, Shiva Raj Pokhrel, and
Jinho Choi are with the School of Information Technology, Deakin University,
Geelong, VIC 3220, Australia (e-mail: yxdj2010@gmail.com, {nematim,
shiva.pokhrel, jinho.choi}@deakin.edu.au).
Ok-Sun Park is with the Electronics and Telecommunications Research
Institute, Daejeon, South Korea (e-mail: ospark@etri.re.kr).
Fumiyuki Adachi is with the Research Organization of Electrical
Communication, Tohoku University, 980-8577 Japan (e-mail:
adachi@ecei.tohoku.ac.jp).
for 5G [3], namely enhanced mobile broadband (eMBB), mas-
sive machine-type communication (mMTC), and ultra-reliable
low-latency communication (URLLC). Among them, enabling
URLLC service is the most challenging task in 5G and future
wireless networks as two ambitious requirements of high
reliability and low-latency need to be satisfied simultaneously
for short-packet transmissions [4]–[6].
A. Background
Emerging URLLC applications include virtual/augmented
reality, public safety, factory automation, and autonomous
vehicles, among others [7]. 3GPP has already identified sev-
eral scenarios for factory automation, where actuation over
industrial devices have stringent performance requirements
(latency of 1 millisecond (ms) and reliability of 99.9999%
[8]). Earlier studies on the long-term evolution (LTE) systems
to support URLLC [9] reported that LTE is not applicable to
meet low-latency requirements. The reason being its typical
uplink radio access delay of 7.5 ms and handover delay of
50 ms. To solve this, a new radio interface called 5G new
radio (NR) has been standardized. In 5G NR, a number of
advances, including waveform numerology, frame structure,
multiple access schemes, and scheduling policy, have been
adopted for enabling URLLC [10]–[12].
Of particular relevance to this work is the novel grant-free
(GF) random access procedure in NR. GF random access is to
overcome the drawbacks of the conventional random access
procedure in LTE, which is now a key enabler to support
uplink URLLC. The traditional idea in LTE follows the grant-
based random access procedure for a device to access channel,
where it needs to request and obtain access grants via four-way
handshaking with a base station (BS) [13]. Such handshaking
ensures that the device has an exclusively reserved channel for
contention-free transmission, avoiding any potential collisions
in data transmissions. However, it incurs additional latency
and undesirable signalling overheads, which hinder achieving
the required level of latency constraints for URLLC. For
this reason, the request-grant handshaking procedure has been
removed in GF to attain prompt channel access (no sending
request and waiting time for channel grant [10]).
In GF URLLC, since multiple competing devices can
transmit data over the same channel, it results in potential
transmission collision, which is detrimental to the reliability.
To alleviate the issue, various hybrid automatic repeat request
(HARQ) retransmission schemes have been considered in GF
URLLC in academic research and standardization bodies [12],
2
[14], [15]. For instance, except traditional reactive scheme,
the state-of-the-art HARQ retransmission schemes include the
K-repetition scheme, where devices blindly retransmit data
packets multiple times before receiving feedback, and the
proactive scheme, where devices proactively retransmit until
receiving positive feedback. In addition, transmission schemes
such as multi-connectivity [16] and fast retrial [17] have been
proposed to support GF URLLC. Nevertheless, with limited
latency budget and wireless resources, their resulting reliability
levels and spectral efficiency still needs to be enhanced to meet
what is required for emerging URLLC services, particularly
when the URLLC access load is relatively high [6].
In recent years, real momentum has been building up
for deploying and advancing multiple-input multiple-output
(MIMO) to enhance wireless communications’ reliability and
spectral efficiency [18]–[20]. Massive MIMO (mMIMO) [21],
conceived in 2010, has the capability to handle a large number
of antennas at a BS to serve IoT devices simultaneously.
It has now been widely investigated and helped enhance
several areas of wireless communications [22]. Thanks to
mMIMO advances, large spatial diversity and multiplexing
gains coupled with a large array gain can now be realized
seamlessly. In addition, such advancements of mMIMO can
be exploited to significantly enhance transmission reliability
and spectral efficiency along with other critical performance
metrics [23]–[28]. Thus, mMIMO has the potential for a
substantial impact on GF random access in supporting a large
number of URLLC devices [29]–[31].
Looking further ahead, a new distributed mMIMO architec-
ture, so-called cell-free (CF) mMIMO, has now attracted a lot
of attention from researchers and industry verticals [32]–[34].
CF mMIMO can be one of the key enablers for 6G wireless
networks [35]. In a sharp contrast with (centralized) mMIMO
(where co-located antennas are deployed in a compact area),
antennas are spread out in CF mMIMO over a large area to
serve devices without the notion of cell boundaries [32]. Since
CF mMIMO reaps all the benefits from mMIMO as well as
network MIMO [36], we anticipate that CF mMIMO will open
up new avenues and advances for GF URLLC.
In the literature, there are several survey and tutorial
papers on URLLC and mMIMO technology. For instance,
a comprehensive overview of the physical layer design in
NR for URLLC, including fundamental changes compared to
LTE systems, can be found in [37]. The building principles
of URLLC at the high layer were highlighted in [38]. A
review of challenges and approaches for enabling URLLC
within mMTC was provided in [3]. In [4], the principles of
access protocols for URLLC and the fundamental tradeoffs
from a communication-theoretic perspective were discussed.
In [5], the authors focused on the study of URLLC at the
network layer, where some of the key enablers of URLLC
were introduced, and various mathematical tools tailored to the
unique features of URLLC were examined. In [39], a tutorial
on how to combine theoretical knowledge and deep learning
algorithms to optimize URLLC in a cross-layer manner was
investigated. In addition, opportunities and challenges in ex-
ploiting mMIMO for IoT connectivity were identified in [2]
[31] and review studies on CF mMIMO can be found in [33]
[34]. Unlike the existing articles, this survey mainly focuses
on enabling GF URLLC and addressing its fundamental issues
by exploiting mMIMO from the physical and data-link layers
perspective. To the best of our knowledge, there still lacks
a comprehensive investigation and insight into the state-of-
the-art and potential opportunities for enabling GF URLLC in
mMIMO/CF mMIMO.
VI. Conclusions
Overview Structure
I. Introduction
• Background
• Contributions
• Organization
III. Enabling URLLC in NR
• Flexible Frame Structure
• GF Random Access
• HARQ Retransmission Schemes for GF URLLC
• Fast Retrial
IV. Significance of mMIMO for GF URLLC
• Features of mMIMO
• Enhancements to GF URLLC
II. Use Cases of URLLC
• Smart Factory & Industrial Automation
• Intelligent Transportation Systems
• Healthcare & Public Safety
V. Potential of Emerging CF mMIMO
• Features of CF mMIMO
• Research Directions and Challenges
Overview of
Key Enablers
Enhancements
by mMIMO
Visions on
CF mMIMO
Fig. 1. Organization of the paper and interactions among different sections.
B. Contributions
As GF URLLC and its enhancements by mMIMO/CF
mMIMO have been attracting more and more attention from
academia and industries, it is worth aggregating the existing
findings and laying a foundation for future research direction
on these subjects. With this as a primary objective, our three
main contributions in this paper are outlined as follows:
1) We develop a comprehensive review of NR specifica-
tions and techniques for URLLC, discussing underlying
principles and highlighting fundamental issues of en-
abling URLLC with GF random access.
2) We review the key phenomena of mMIMO and build
several deep insights into how mMIMO can be exploited
to address the fundamental issues and enhance the
performance of GF URLLC. In particular, we explain the
benefits of exploiting preamble-collision information,
coded random access, and multi-preamble detection for
GF URLLC, which are only achievable via mMIMO.
3) We consider the potential of CF mMIMO and examine
its distinctive features and benefits over centralized
mMIMO to address impeding GF URLLC bottlenecks.
Based on our understanding, we project future research
directions and challenges of enabling GF URLLC in CF
mMIMO.
3
C. Organization
A high-level simplified view of the paper structure is shown
in Fig. 1. In Section II, we provide an overview of typical
5G URLLC use cases, highlighting the benefits of adopting
GF random access and mMIMO to enable these use cases.
To better understand the main enablers of URLLC, Section
III provides a comprehensive review of the principle of key
enablers in 5G NR, including GF random access, and identifies
fundamental issues of enabling GF URLLC. Consequently, in
Section IV, we do an in-depth investigation on the significance
of mMIMO for GF URLLC and how the combination of these
techniques is applied to help address the identified issues.
Furthermore, Section V provides insights into future research
directions focusing on the potentials of CF mMIMO for GF
URLLC. Lastly, Section VI concludes the paper.
II. USE CASES OF URLLC
URLLC use cases set stringent transmission requirements,
e.g., 99.999% transmission success rate within 1 ms [40].
We herein focus on the three main real-life 5G business use
cases: i) Industry 4.0 smart manufacturing [41], ii) Connected
autonomous vehicles [42], and iii) robotic surgeries [43], all
of which are now in the deployment phase. Since GF random
access has been a part of 5G standards, it is worth noting
that low delay and time-sensitive network applications (e.g.,
automated car driving and near real-time robotics) of the three
use cases can benefit significantly with the GF URLLC [44].
Refer to [3], [7], [40], [44]–[48] and references therein to see
other relevant URLLC use cases in different industry verticals.
Significant trends in such applications are user-specific
three-dimensional (3D) video rendering, augmented reality,
remote control (e.g. remote robotics, surgery, tactile internet,
etc.), wireless communication automation for efficient produc-
tion facilities, vehicular traffic efficiency/safety, and mobile
gaming, among others. As shown in Fig. 2, we briefly review
three applications of URLLC along with their requirements in
the following.
A. Smart Factory and Industrial Automation
Smart factory as part of industrial automation is one of
the key applications demanding URLLC. This is reflected, for
example, in the Industry 4.0 migration, the process control
mechanisms have been automated and deployed using the
advances in wireless networks [41]. Indeed, Industry 4.0
deployment includes different ranges of real-time interface and
actuation such as: process automation, motion control, indus-
trial Ethernet, power system automation, and other control-
cum-communication requirements [47], [49]. To enable factory
automation, it requires real-time interactions among multiple
machines, devices, robots and plants. In addition, it is worth
noting that missing a deadline can be very costly as the
automation systems are always cascaded and characterized by
the need to meet their inter- and intra-system deadlines [50].
Furthermore, ultra-tight synchronization should be consid-
ered to be the third axis of URLLC when targeting such
critical use cases for industrial automation [46]. For instance,
synchronization accuracy of about 5µs is required for fault
location identification [7]. The latency requirement becomes
even tighter in automation processing monitoring and motion
control with 50 ms and 1 ms along with the reliability of
99.9% and 99.9999%, respectively [45].
Recently, 5G NR has been considered an appropriate in-
frastructure for real-time interactions among such industrial
entities. In [51], GF random access in 5G NR with priority-
based schemes was studied in an industry automation scenario.
It is shown that the GF schemes outperform legacy ones in
supporting Industry 4.0 requirements. In addition, [52] studied
the performance of GF URLLC in a factories-of-the-future
scenario and indicated that using spatial diversity techniques
appears to be the most suited strategy to achieve ultra-reliable
GF transmission within a target latency.
B. Intelligent Transportation Systems
Intelligent transportation systems (ITS) is another impor-
tant application of URLLC, which enables communications
between transportation entities, including different types of ve-
hicles, unmanned aerial vehicles (UAVs), trains, roadside units
etc. One prevalent example is the Internet of Vehicles (IoV),
wherein on-vehicle distributed learning models are trained by
exchanging inputs, outputs, and their learning parameters in an
ad hoc fashion [53]. To be specific, the latency requirement for
such distributed learning and connections of the ITS entities
with the infrastructure backhaul should be lower than 10
ms, along with a reliability of 99.9999% [45]. For example,
missing deadlines in applications of autonomous vehicles can
be highly catastrophic [42]. Also, we are still in an early
phase of ITS deployment, and we have been suffering from
limitations in case of emergencies which results in larger
delays in broadcasting emergency messages [54], [55]. Such
dissemination demands URLLC and becomes more critical in
the case of high-speed trains and aerial vehicles [56]. Besides,
for vehicle to vehicle (V2V) communication in connected
autonomous vehicles, the latency requirement of 1 ms is
mandatory. Simultaneously, with the current IEEE 802.11p
protocol, V2V communication suffers from unbounded latency
and highly varying reliability [53], [57].
Motivated by these, GF random access has been considered
and identified as a key enabler for 5G empowered ITS [44].
Furthermore, as indicated in [58], the emerging mMIMO
technology can facilitate gigabits-per-second (Gbps) commu-
nication for a variety of cellular ITS scenarios, which is far
beyond what the legacy dedicated short-range communication
(DSRC) [59] and LTE-advanced systems can support.
C. Healthcare and Public Safety
Health care applications such as robotic telesurgery and
tactile Internet spur the need for URLLC. The remote surgical
consultations and remote surgery, so-called telesurgery, adopt
augmented reality (AR) and virtual reality (VR) advancements.
For instance, surgeons use VR headsets to observe/actuate the
inside of a patient. They also require patients over VR headsets
when taking them through their surgical plan [47]. The con-
nectivity requirements for the explosive growth of such devices
and systems with sensor-based implementations in healthcare
4
Smart Factory and
Industrial Automation
Intelligent Transportation
Systems
Healthcare and
Public Safety
Cyber
security
Vehicle-to-vehicle
Maritime transport
Virtual reality
Augmented reality
Aerial vehicles
Autonomous
cars
Virtual reality
Augmente
High-speed trains
Motion control
3D printing
Manufacturing
Process control
Telesurgery
Robot
hand
Ventricular
Fig. 2. Three real life use cases of URLLC. Among others, the significance of URLLC in these use cases can be observed in terms of the following three
examples: i) aids in the automation of manufacturing operations and factory control systems; ii) facilitates surgical augmented-reality which allows surgeons
to undertake surgical resection in a much more precise and analytical manner, reducing the risk of relapse; iii) enables route discovery and collision avoidance
in real-time relying heavily on communications involving cars and everything else in the transportation grid.
facilities will accelerate the growth of mMTC. The required
end-to-end latency for such application of health care verticals
are low, ranging from 125 ms to 1 ms (for mission-critical
healthcare applications) with 99.9999% reliability guarantees
[43].
Likewise, public safety requires robust and reliable com-
munications in case of natural disasters such as earthquakes,
tsunamis, floods and hurricanes [60]. Specifically, accurate po-
sitioning, fast communications, real-time video, and the ability
to send high-quality pictures in the presence of damaged wired
technologies are critical challenges for public safety [49].
Various studies have been carried out to enable applications
in both domains. For instance, [61] advocated the use of
GF and non-orthogonal transmissions to achieve a significant
reliability gain for tactile Internet. And [62] reviewed emerging
technologies for enabling public safety services, and mMIMO
is one of them.
Overall, emerging URLLC use cases are posing unprece-
dented challenges in terms of latency, reliability, and scalabil-
ity for systematic design [63]. As reviewed, GF random access
and mMIMO have the potential to address the challenges
[31], [44]. To provide a comprehensive and insightful view,
in the following, we first present the principles of enabling
URLLC in 5G NR, including GF random access and identify
fundamental issues of enabling GF URLLC. Based on them,
we then shed light on the potential of ameliorating the issues
and advancing URLLC capabilities with the assistance of
mMIMO technology, which can be served as guidelines for
underpinning the applications of URLLC.
III. ENABLING URLLC IN NR
To meet latency and reliability requirements of various
URLLC use cases, key enabling techniques in NR physical
and medium access control (PHY/MAC) layers have been
adopted in 3GPP release 15 and enhanced in 3GPP releases
16 and 17 [11], [49], [64]. In this section, we explain several
key NR specifications and techniques for URLLC, including
flexible frame structure, GF random access, and retransmission
schemes.
A. Flexible Frame Structure
In NR, one of the main specifications adopted for URLLC is
a flexible frame structure, which is able to not only reduce la-
tency but also create more retransmission opportunities within
a target latency that in turn lead to enhanced reliability [45].
Since the user-plane, latency
1
is one of the dominant com-
ponents in the latency of URLLC and the transmission time
interval (TTI) plays an important role in contributing to the
user plane latency, reducing TTI is a key to meet the low-
latency requirement. For this reason, NR enables TTI reduction
by introducing scalable numerology (subcarrier spacing) and
the concept of mini-slots [49].
Specifically, the subcarrier spacing in LTE is fixed at 15
KHz, and the basic TTI is set to 1 ms, which equals the length
of a subframe/slot. Different from it, the TTI of URLLC can
be shortened by increasing the subcarrier spacing and/or using
small scheduling units such as mini-slots. As illustrated in Fig.
3, the subcarrier spacing of 2
n
× 15 KHz can be configured
for URLLC data transmissions depending on bandwidth, de-
ployment, and use cases, where n can be 0, 1, or 2 in Sub-6
GHz and 3 in the millimeter-wave spectrum. Since a larger
subcarrier spacing can be employed than that of the baseline of
15 KHz, the duration of slot and orthogonal frequency division
multiplexing (OFDM) symbol can be significantly reduced.
For instance, the slot-based TTI with 60 kHz subcarrier
spacing is 0.25 ms, which is only a quarter of that with 15
KHz.
On top of this, the number of OFDM symbols in each TTI
does not necessarily equal 14. To be more precise, a mini-slot
can be employed to shorten further the TTI, consisting of 1
to 13 OFDM symbols. In Fig. 3, a mini-slot based TTI of 2
1
User-plane latency is the time it takes to successfully deliver a data packet
at the radio protocol layer from the transmitter to the receiver.
5
15KHz
Subframe of 1ms
30KHz Slot
Mini-Slot
60KHz Slot Slot
Mini-Slot
One OFDM Symbol
Fig. 3. Illustration of slot and mini-slot structure for different numerologies
[49].
OFDM symbols with the subcarrier spacing of 30 KHz and
a mini-slot based TTI of 7 OFDM symbols with 60 KHz are
illustrated, respectively. Evidently, with a subcarrier spacing
of 30 KHz, a 70 µs TTI is achieved by a 2-symbol mini-slot
as opposed to 0.5 ms based on slot-based transmission. This
demonstrates a substantial latency reduction by adopting short
TTI. Furthermore, suppose a target latency of 1 ms, when the
mini-slot is adopted, 14 transmission opportunities are created
within the latency, which can be exploited to enhance the
transmission reliability.
B. GF Random Access
To implement a radio access network for URLLC, handling
random access design can be one of the most important/critical
portions. GF random access (also known as 2-step random
access the [65]) procedure is another key NR specification to
enable URLLC.
BS
Device
Scheduled Transmission
Contention Resolution
Random Access Response
BS
Device
MsgA: Preamble and Data
MsgB: Random Access Response
Msg1 + Msg3
Msg2 + Msg4
Random Access Preamble
Msg1
Msg2
Msg3
Msg4
4-Step Random Access
2-Step/GF Random Access
Fig. 4. Illustration of GF random access procedure compared to conventional
4-step random access procedure [11].
In LTE, a 4-step random access procedure, also known
as grant-based random access, is adopted, which is mainly
designed for human-type communication and not suitable for
MTC, including URLLC [9], [66]–[68]. As illustrated in Fig.
4, in the first step of the 4-step random access, each active
device transmits a randomly selected preamble on physical
random access channel (PRACH) to initiate a scheduling
request. In the second step, the BS detects the preambles
transmitted by active devices and sends responses by issuing a
scheduling grant. Once an active device is connected to the BS,
and it can transmit data packets in the third step on dedicated
resource blocks (RBs) or channels, which are physical uplink
shared channel (PUSCH). The 4-step random-access procedure
requires two round-trip cycles between the devices and the
BS, which raises the barriers to meet the stringent latency
requirement of URLLC use cases [9]. In particular, it not only
increases the latency but also incurs large control-signaling
overhead for small packets [11].
The motivation of GF random access is to reduce latency
and control-signaling overhead by having a single round-trip
cycle between the devices and the BS. Compared to the 4-step
random access, GF random access can be more efficient thanks
to low signaling overhead when devices have short packets to
transmit. In GF random access, an active device does not wait
for a scheduling grant from the BS. That is, once a device
becomes active in a random-access slot (TTI), it is to transmit
a preamble directly along with data on the same channel in a
time division multiplexing (TDM) manner and waits for the
acknowledgement from the BS.
In GF random access, contention-free and contention-based
transmission modes can be operated. In this paper, we mainly
focus on the latter transmission mode. The contention-free
GF transmission can be employed when the number of active
devices is small and their access traffic is periodic or determin-
istic. It allows the BS to pre-allocate wireless resources, e.g.,
preamble and spectrum resources, to devices so that access
contention can be avoided [69], [70]. In the case of sporadic
traffic or/and massive URLLC, contention-based GF transmis-
sion is more suitable since it is more efficient and flexible
in terms of resource utilization [71]. Nevertheless, contention-
based GF transmission is prone to potential access collisions
when multiple devices simultaneously access the same channel
resource, thus jeopardizing transmission reliability [52], [71],
[72].
C. HARQ Retransmission Schemes for GF URLLC
There are a number of techniques to improve transmission
reliability. We briefly discuss HARQ retransmission schemes
for URLLC in this subsection. Thanks to the flexible frame
structure in NR, as mentioned above, multiple retransmission
opportunities can be created within a target latency by adopting
short TTI. As a result, several different HARQ retransmission
schemes can be incorporated into GF random access to support
URLLC.
1) Reactive HARQ Retransmission Scheme: Reactive
HARQ retransmission scheme is a conventional scheme
adopted in LTE. In this scheme, retransmission is allowed
only when a device receives a Negative ACKnowledgement
(NACK). Once the device transmits a data packet, the device
has to wait for feedback from a BS before any retransmission
attempt. To issue the feedback, the BS is to receive and process
the data packet from the device. In particular, as shown in
Fig. 5(a), with the assumption that both the BS and the device
spend 1 TTI for processing and 1 TTI for transmitting [74],
we see that the HARQ round-trip-time (RTT
2
) takes 4 TTI in
the scheme, which means that the device needs to wait for 4
TTI until the next retransmission attempt. In this case, if the
2
RTT is the time duration of the cycle from the beginning of transmission
until processing its received feedback [73].
