Empowering the 6G Cellular Architecture
with Open RAN

Michele Polese, Mischa Dohler, Falko Dressler, Melike Erol-Kantarci,
Rittwik Jana, Raymond Knopp, Tommaso Melodia

Abstract—Innovation and standardization in 5G have brought
advancements to every facet of the cellular architecture. This
ranges from the introduction of new frequency bands and signaling technologies for the radio access network (RAN), to a core
network underpinned by micro-services and network function
virtualization (NFV). However, like any emerging technology, the
pace of real-world deployments does not instantly match the pace
of innovation. To address this discrepancy, one of the key aspects
under continuous development is the RAN with the aim of making
it more open, adaptive, functional, and easy to manage.
In this paper, we highlight the transformative potential of

it more open, adaptive, functional, and easy to manage.
In this paper, we highlight the transformative potential of
embracing novel cellular architectures by transitioning from
conventional systems to the progressive principles of Open RAN.
This promises to make 6G networks more agile, cost-effective,
energy-efficient, and resilient. It opens up a plethora of novel use
cases, ranging from ubiquitous support for autonomous devices
to cost-effective expansions in regions previously underserved.
The principles of Open RAN encompass: (i) a disaggregated
architecture with modular and standardized interfaces; (ii)
cloudification, programmability and orchestration; and (iii) AIenabled data-centric closed-loop control and automation. We first
discuss the transformative role Open RAN principles have played
in the 5G era. Then, we adopt a system-level approach and
describe how these Open RAN principles will support 6G RAN
and architecture innovation. We qualitatively discuss potential
performance gains that Open RAN principles yield for specific 6G
use cases. For each principle, we outline the steps that research,
development and standardization communities ought to take to
make Open RAN principles central to next-generation cellular
network designs.

I. INTRODUCTION

The wireless internet plays a fundamental role today in
supporting societies and economies around the world, underpinned by cellular connectivity that is widely used by
consumers but also in industries, health, education, and entertainment. This is testament to the revolution that fourth
and fifth generations (4G and 5G) of cellular networks have

introduced, making it easy to stream data at high rates to
mobile phones, to provide connectivity to vehicles and sensors,
and so much more.
More than ten years of releases from 3GPP have defined

More than ten years of releases from 3GPP have defined
the Long Term Evolution (LTE) and NR technologies underpinning the 4G and 5G radio access network (RAN) and core
network \[1\]. These cellular systems share an IP-based core network design, an Orthogonal Frequency Division Multiplexing
(OFDM) waveform in the RAN physical layer, end-to-end data
abstractions to apply security and Quality of Service (QoS) to
user flows, and robust mobility procedures, among many other
novel principles. 5G further advances cellular systems with
an array of first-of-its-kind solutions, including support for
communications in the lower portion of the millimeter wave
(mmWave) band \[2\], directional communications \[3\], massive
Multiple Input, Multiple Output (MIMO) \[4\], and a frame
structure which supports different user traffic requirements \[5\].
Nonetheless, while the standard specifications feature sev-

structure which supports different user traffic requirements \[5\].
Nonetheless, while the standard specifications feature several relevant techniques to provide data rates, latency, and other
Key Performance Measurements (KPMs) in line with the 5G
definitions from the International Telecommunication Union
(ITU) \[6\], actual 5G deployments lag behind the standardization by at least one release: indeed, the 5G Releases 16 and 17
have already been finalized whilst only a subset of Release 16
features is commonly deployed, and often in non-standalone
mode \[7\]. In addition, the usage of the mmWave spectrum
has not taken off \[8\]. Finally, today’s 5G networks are largely
deployed using global default configurations, neglecting the
potential optimization benefits available with bespoke and
tailored setups \[9, 10\].
This is because of several reasons, however, financial being

F. Dressler is with the School of Electrical Engineering and Computer
Science, TU Berlin, Germany. Email: [dressler@ccs-labs.org](mailto:dressler@ccs-labs.org).
M. Erol-Kantarci is with the School of Electrical Engineering and Com-

Science, TU Berlin, Germany. Email: [dressler@ccs-labs.org](mailto:dressler@ccs-labs.org).
M. Erol-Kantarci is with the School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, ON, Canada. E-mail: [melike.erolkantarci@uottawa.ca](mailto:melike.erolkantarci@uottawa.ca).
R. Jana is with Google, NYC, NY. E-mail: [rittwikj@google.com](mailto:rittwikj@google.com).
R. Knopp is with Eurecom, Sophia Antipolis. E-mail: ray-

This is because of several reasons, however, financial being
the main one. And whilst the largest bottom-line item on the
balance sheet for cellular operator remains customer churn
and the largest unrealized top-line item remains the inability
to charge for innovative 5G use-cases, an important cost factor
remains: the high deployment and operational costs that are
associated to installing, configuring, optimizing, and operating
5G equipment without a clear business strategy for the return
on investment. While the research and development ecosystem
has limited control on the business and market development,
we believe that there are technical, architectural, and systemlevel solutions that, if natively integrated in the design of 6th
generation (6G) networks, can kickstart a faster innovation
cycle for cellular, introducing approaches that are widely
adopted in cloud environments and the software industry.
Discussions are currently well underway on what kind of

This work was partially supported by the U.S. National Science Foundation
under Grant CNS-2117814 as well as by the Federal Ministry of Education and
Research (BMBF, Germany) within the 6G Platform under Grant 16KISK050.

* * *

The recent literature, however, lacks a vision on how we can
address fundamental issues related to the network architecture
and introduce a clean-slate, system-level re-design of nextgeneration cellular networks. This paper fills this gap by
providing a systematic review of the architectural limitations
of existing cellular networks and deployments, and and a
tutorial on how the foundational principles of Open RAN can
be embedded into the 6G cellular architecture to address some
of such limitations.
Notably, we envision 6G networks to be data-driven, au-

of such limitations.
Notably, we envision 6G networks to be data-driven, autonomous systems where algorithmic control and Artificial
Intelligence (AI) are systematically applied to fine-tune and
optimize programmable protocol stacks. Further, virtualized
(and possibly disaggregated) network functions are dynamically and automatically orchestrated to comply with specific
performance requirements of use cases and application. These
networks are primarily deployed through software, with hardware acceleration, and can make an agile use of dedicated or
shared spectrum, compute, and infrastructure resources. They
can be deployed with minimal effort, and can extend coverage through automated and optimzed self-backhaul solutions.
Finally, Open-RAN-based 6G systems can leverage the same
principles of programmability, automation, and virtualization
to adopt cloud-native zero-trust security and resiliency strategies. The innovation introduced by Open RAN in wireless
systems is comparable to the transformation that Softwaredefined Networking (SDN) and programmable user planes
have brought along in Ethernet switching systems \[17, 18\],
opening opportunities for closed-loop and intelligent algorithmic control of once inflexible platforms.
The main contribution of this paper is to clearly articu-

mic control of once inflexible platforms.
The main contribution of this paper is to clearly articulate and describe the key constituents that are required to
implement this vision, providing a tutorial that can illustrate
and direct research efforts toward Open RAN in 6G. Compared to \[9, 16, 19–26\], which mostly focus on the building
blocks and the development of Open RAN systems, in this
paper we analyze cellular networks as complex systems and
identify opportunities towards innovation at the system level.
The second contribution of this paper is connecting key 6G
performance requirements (e.g., energy efficiency, ubiquitous
coverage, resiliency, low cost and complexity) to Open RAN
principles, discussing the performance and operational gains
that open, virtualized, and intelligent networks enable. Finally,
for each of the building blocks of Open RAN systems, we
discuss current state of the art and research directions.
Overall, this paper provides a map and a vision to steer

Fig. 1: Components and abstractions of 4G and 5G cellular networks, with
the limitations associated to the current cellular architecture and deployment
models.

II. 4G AND 5G CELLULAR NETWORK ARCHITECTURE

The design of cellular networks separates functionalities
across different elements—the RAN and core network, in a
horizontal split—and planes—the user and the control planes
(UP and CP), in a vertical split \[27\]. They support connectivity
in mobile scenarios and provide exclusive spectrum access
to licensed and shared bands through centralized scheduling.
Compared to other popular wireless technologies such as Wi-
Fi, cellular networks support a more widespread coverage
layer with a consistent experience also in mobile scenario, and
can extend performance guarantees to the end users. However,
existing deployments still present limitations, e.g., they come
with limited reconfigurability and adaptability to use cases and
users’ requirements and rely partly on manual optimization.
In this section, we review the high-level characteristics

users’ requirements and rely partly on manual optimization.
In this section, we review the high-level characteristics
of cellular network architectures that have been typically
deployed in 4G and 5G networks, primarily with 3GPP LTE
and NR \[1, 27\], and discuss opportunities for the ecosystem
to accelerate the pace of innovation.

The first functional split in wireless networks is horizontal,
i.e., two different parts of the system provide last-mile access
to the users (i.e., the RAN) and support functionalities (i.e.,
the core network).
As discussed in \[1, 5\], the RAN comprises the set of base

the core network).
As discussed in \[1, 5\], the RAN comprises the set of base
stations, either Next Generation Node Bases (gNBs) for a
5G NR RAN, or evolved Node Bases (eNBs) for a 4G LTE
RAN. The base stations provide wireless access to the mobile
network to User Equipments (UEs); in 4G, they are generally
deployed as monolithic units, and in 5G as virtualized (vRAN)
and even cloudified (C-RAN) instantiations. As shown in
Fig. 1, each base station is equipped with a complex protocol
stack defined by the 3GPP, where the physical layer takes

A. Access and Core Network care of Digital Signal Processing (DSP), channel estimation,
transmission and reception, and the data link layer, in charge
of scheduling and QoS optimization, is broken down into
multiple sub-layers. The Medium Access Control (MAC) layer
handles scheduling; the Radio Link Control (RLC) layer
buffering, concatenation, and segmentation; the Packet Data
Convergence Protocol (PDCP) layer performs data encryption
and packet sequencing; the Service Data Adaptation Protocol
(SDAP) layer enforces QoS; and the Radio Resource Control
(RRC) layer implements the state machine of the network,
among other things \[27\]. This stack has evolved across different generations of 3GPP specifications, e.g., the SDAP layer
has been introduced in NR for 5G systems.
The core network primarily manages user authentication,

The core network primarily manages user authentication,
mobility, paging, and routing to and from the public Internet.
Fig. 1 reports a high-level illustration of the the 4G core,
or Evolved Packet Core \[28\], and of the 5G core, which is
based on chaining and orchestrating multiple atomic network
functions through a Service-based Architecture (SBA) \[29\].
Whilst the 4G core is typically deployed via monolithic
servers, the 5G core is fully virtualized and can be deployed
via virtual machines or Kubernetes-based micro services. The
SBA approach, however, has not yet fully realized its potential.

B. User and Control Plane

The second functional split is vertical, i.e., there are different
procedures, protocols, layers, and core network components
to manage the transmission and reception of user data (user
plane) and the connectivity lifecycle (control plane). This separation introduces a clean distinction between configuration,
optimization, and management, which are under the purview
of the control plane, and the processing of the data in any
shape (e.g., bits, packets), which is handled by the user plane.
The user plane processes data hierarchically. First, it man-

C. Limitations and Theory-Implementation Gap

The user plane processes data hierarchically. First, it manages IP packets, for example by applying encryption at the
PDCP layer and associating QoS levels to end-to-end packet
streams, or bearers, at the SDAP layer. Then, packets buffered
at the PDCP and RLC layers, which segment and concatenate
them at a byte level into transport blocks, based on the
resources scheduled by the MAC. The physical layer then
transforms them into a waveform by processing the input at
a bit level. These operations are performed in reverse order
at the receiver. In the core network, specific functions (e.g.,
the User Plane Function, UPF, in the 5G core) act as packet
gateways between the public Internet and the telecom operator
network, and establish bearers with the mobile devices.
The control plane guides the components of the user plane to

These cellular network design principles are well-thought
and the outcome of years of evolution of cellular systems, with
5G networks reaching hundreds of Mbps of throughput in typical scenarios \[8, 30, 31\]. However, there is still a gap between
cellular deployments and the rich set of features available in
the specifications from 3GPP and in the technical literature
from the wireless and networking research communities. The
capabilities of cellular networks are today mostly focused on
broadband applications, and lag behind the vision of what
applications could achieve in mobile scenarios:

• First, the complexity, scale and performance requirements
as well as lack of a clear monetizable use-case of
production-grade cellular networks lead operators and
thus vendors to focus on a subset of well-tested features
to mainly support mobile broadband applications. This
limits the potential of cellular network deployments, as
well as the number of parameters that can be tuned and
optimized in production. A further setback is that very
few operators have upgraded to 5G standalone networks,
complicating the support for different kinds of use cases
at the same time, e.g., ultra-low latency data streaming for
industrial control on a public network while also serving
video streaming to residential users.
• Second, the limited flexibility extends to spectrum, the

• Second, the limited flexibility extends to spectrum, the
most valuable resource when deploying wireless systems.
Cellular networks, so far, have been mostly deployed
on licensed spectrum, which provides guarantees for no
or limited interference, but also limits the bandwidth to
small chunks and disaggregated chunks of spectrum (at
least below 6 GHz). More flexible radios and spectrum
access systems can pave the way for more efficient spectrum utilization and an overall increase in the available
bandwidth.
• Third, there are limited options for the deployment of the

• Third, there are limited options for the deployment of the
network itself. Whilst 3GPP interfaces and the standard
itself are completely open, market consolidation has led
to a small set of RAN vendors. Whilst vendor interoperability is ensured via well-tested Xn interfaces, further
disaggregation could potentially spurn more innovation
and also increase the resiliency of the supply chain \[32–\
34\].
• Finally, whilst the 5G architecture allows for cloudified

Overall, complexity, lack of adaptation, and limited flexibility offset the benefits introduced by the robust protocol
stack design of cellular systems, and potentially prevent the
adoption of state-of-the-art techniques to provide more innovative user services. This challenges the emergence of new
market entrants for private and public cellular systems, limiting
the diversity of the supply chain and telecom ecosystem and
stymieing fast-paced competition and innovation.

* * *

Fig. 2: Foundational architectural principles of next-generation cellular networks (black), overlayed on the Open RAN architecture (orange), and with 6G
target Key Performance Indicators (KPIs) (blue) related to energy efficiency, cost and complexity, innovation, coverage, and resiliency.

III. EMPOWERING THE 6G CELLULAR ARCHITECTURE
WITH OPEN RAN PRINCIPLES

• Energy efficiency — energy consumption is one the main
cost drivers for telecom operators (up to 60% of the
operational expenses \[35\]), and, overall, the information
and computing technologies industry contributes to two to
three percent of the total greenhouse gases emission \[36\].
To this end, recent literature has focused on improving the
energy efficiency as one of the key elements of the design
of next-generation cellular systems \[37–39\], improving
radio components, protocols, efficiency, and utilization.
Lower deployment and operational costs and com-

In this section, we describe how 6G can benefit from
the adoption of a new, system-level approach for the design and deployment of the network architecture, leading
to a streamlined transition of innovation from research and
standardization to production. It is important to distinguish
between standards for key functionalities (as mainly done by
the 3GPP); for enabling a softwarized/cloudified deployment
with closed-loop control (e.g., O-RAN Alliance); and the
actual deployments.
The quest that led the development of 4G and 5G networks

radio components, protocols, efficiency, and utilization.
• Lower deployment and operational costs and complexity — upgrading cellular networks across generations
is a complex and costly exercise for network operators,

with cycles that take years to complete (e.g., one major
U.S. operator moved commercial traffic to a 5G standalone core network in 2022, three years after launching
a commercial 5G service \[40\]). Similarly, operations
require careful planning and supervision with humans
in the loop, increasing expenses and limiting the scope
of optimization to few selected parameters and services.
Future wireless systems will need streamlined and automated processes for network deployment and operations.
• Faster innovation cycles — once a network is opera-

systems and reducing cost for deployment \[41–43\].
High resiliency — the cellular infrastructure is critical

mated processes for network deployment and operations.
• Faster innovation cycles — once a network is operational, deploying and testing new features and functionalities, e.g., features from new 3GPP releases, represents
an additional effort and cost, as it could lead to Service
Level Agreement (SLA) requirements violation and risk
of service disruption. This slows down innovation and
prevents a quick adoption of new techniques and solutions
in cellular networks—a key challenge to address in nextgeneration networks to unlock faster transition from lab
research to cutting-edge commercial products.
• Ubiquitous coverage — the deployment of highfrequency 5G networks has focused so far on dense urban

* * *

This is in addition to further improvements in throughput
and latency, e.g., through ultra-wide band networks in the subterahertz spectrum \[45, 46\]. To achieve these target, there is an
opportunity to rethink the cellular architecture and operations
with a holistic approach which can provide more gains than
the improvements in the lower layers of the RAN protocol
stack (e.g., the redesign the physical layer with AI \[47, 48\]).

A. 6G requirements and Open RAN principles

Figure 2 illustrates how the KPIs discussed above can be
supported by foundational principles of Open RAN and nextgeneration cellular networks. The figure envisions a softwarebased disaggregated Open RAN system which is scalable and
flexible, automated and optimized, and capable to dynamically
access licensed, unlicensed, and shared spectrum for access
and self-backhaul links.
Specifically, through softwarization, network functions are

Specifically, through softwarization, network functions are
implemented as software and deployed on generic compute
solutions, generally coupled with accelerators for digital signal
processing. Via disaggregation, monolithic network components are split into atomic network functions, enabling easier
access to market at the cost of additional integration testing.
Softwarization and disaggregation can potentially improve
energy efficiency when combined with dynamic scaling of
resources and aggregation of RAN components at a data-center
scale. Softwarization also improves the resiliency of the access
to the spectrum, using, e.g., agile software-defined radios \[49\],
as well as of the network infrastructure. Indeed, as done in 5G
today, software components implementing network functions
can be deployed on generic hardware, which can be easily
replaced, and quickly transferred in case a compute node fails
(e.g., via micro-services \[50\]).
Automation streamlines complex operations through a

Automation streamlines complex operations through a
declarative process where tasks are defined in advance and executed at runtime. It cuts cost and complexity, reducing the need
for manual supervision and control, and decreases the time it
takes to apply configurations, restore services, and operate the
network. The operator can then express high-level intents to
guide the automation framework towards tailored deployments
and configurations \[51\]. Automation also improves resiliency,
when combined with softwarization and programmability, and
coverage, as it simplifies the deployment, operations, and
management of network solutions in remote locations and at
an extended scale. Security, however, remains an important
issue because of the increased attack surface.
Optimization provides configurations for programmable

included as part of the target KPIs to optimize; and the
overall performance of the system, thus eventually reducing
the cost per bit. Programmability allows dynamic adaptation
of RAN, compute, cloud, and backhaul networking functionalities through closed-loop control and standardized Application
Programming Interfaces (APIs). To this end, it fosters innovation and improves network resiliency enabling new control
routines and adaptation strategies.
Through spectrum sharing, next-generation cellular net-

Through spectrum sharing, next-generation cellular networks will provide a more flexible air interface and spectrum
access mechanism, which can be designed in different ways
(e.g., access to new bands \[52\], unlicensed cellular \[53\], sharing of cellular bands across different operators \[54\], the neutral
host model \[55\]). This is associated with improved coverage,
as spectrum can be allocated more efficiently based on demand
and availability \[56\], and non-exclusive access to spectrum can
create incentives for operators to deploy networks in remote
or rural locations through reduced spectrum cost. In addition,
non-exclusive access to the spectrum can enable new players
and use cases, as for private 5G deployments, thus increasing
competition in the cellular market.
With self-backhaul, access and backhaul are multiplexed on

competition in the cellular market.
With self-backhaul, access and backhaul are multiplexed on
the same waveform, protocol stack, and portion of the spectrum. It is another ingredient toward improving coverage, as
it allows more flexible network topologies which are not constrained to the availability of fiber connectivity to the access
nodes. Potentially, this can extend to network domains that
are not traditionally considered in cellular systems, as Nonterrestrial Networks (NTNs), thus further improving global
coverage. Wireless self-backhaul, in addition, can extend to
additional portions of the network if compared to traditional
inter-gNB or gNB-to-core backhaul, e.g., for interfaces across
layers of the same gNB, gNB to edge deployments, or across
different radio access technologies (e.g., Wi-Fi and cellular).
How would an Open RAN 6G deployment look like? Let

different radio access technologies (e.g., Wi-Fi and cellular).
How would an Open RAN 6G deployment look like? Let
us consider an example. In a rural location, two small operators
deploy low-cost radios in the field, with small footprint and
energy consumption. These are connected to a local edge
data center through wireless backhaul, where base station
components execute on generic compute resources. The two
operators dynamically share the same portion of the spectrum,
to improve coverage and availability of service to their users,
and leverage slicing and other optimization to serve users with
extremely heterogeneous requirements in mobile scenarios.
The RANs are connected to a remote core (running on a public
cloud shared with other operators) through wireless backhaul.
A centralized management and orchestration plane provides
resiliency, optimization, and high service availability. The
network self-adapts to traffic and usage patterns to minimize
the energy consumption, e.g., adapting transmit power, number
of network functions being executed, compute infrastructure
actively used, and spectrum access.
This, and other use cases with higher impact, are not well

* * *

Fig. 3: Representation of a cellular system where a virtualization layer enables
software-based, programmable network functions.

behind this architectural shift for cellular networks.

IV. ENABLERS: SOFTWARIZATION, PROGRAMMABILITY,
AND VIRTUALIZATION

The transition to software-based, programmable, and virtualized environments is leading cellular networks into more
flexible and dynamically tunable systems, which can benefit
from fast deployment and reprogrammability cycles compared
to traditional cellular deployment.
Softwarization is already mature in the wired Ethernet

Softwarization is already mature in the wired Ethernet
switching domain, where white-label switches can run different flavors of operating systems and programmable protocol
stacks. This transition was spearheaded by seminal papers
published since 2008, e.g., on OpenFlow \[17\], which introduced the concept of SDN and separation of control and
switching for Ethernet campus networks; on operating systems
for networks \[57\]; and on programmable data planes based on
the P4 language \[18\].
Cellular networks have been on the same evolution path

the P4 language \[18\].
Cellular networks have been on the same evolution path
since the introduction of the 5G Core, which, as discussed
in Sec. II, adopts a softwarized service-based architecture. This
has opened the door to multiple open-source and softwareonly implementations of the core network itself \[58–60\], and
softwarization is now being considered for the RAN of nextgeneration wireless systems. The community, however, has
also realized that many of the core network functions only
serve one or a very limited purpose, thus defying the need
of an SBA. Clearly, more work is needed in 6G to ensure a
proper way to disaggregate the functions.
The softwarization of the RAN has its roots in SDN, on

the tight coupling between software and hardware, as shown
in Fig. 3. This presents several advantages, including the programmability of networks, as it becomes easier to update configuration of cellular networks which are not anymore bound
to hardware parameters and constraints, but implemented as
software-based network functions. Other advantages include
the diversification of the supply chain ecosystem in wireless
networks, as software components have a lower barrier of entry
if compared to hardware implementations of a protocol stack.
It also decreases the time from idea development to prototyping and implementation, as transitioning software across
domains is a streamlined process if compared to developing
hardware systems. Further, a software-based approach allows
adopting best practices from the software industry, including
cloud compute solutions, micro-services for resilient and easily
scalable platforms, and fast deployment cycles with quality
guaranteed by a continuous testing, integration and deployment cycle, as we discuss in Sec. VI.
On the other hand, softwarization, as discussed above,

On the other hand, softwarization, as discussed above,
introduces performance challenges related to the implementation and performance tuning of DSP routines. This has led
to (i) effort to define minimum and desirable performance
requirements for whitebox hardware that supports Open RAN
deployments; and to (ii) the emergence of multiple programmable solutions for DSP acceleration, including Graphics Processing Units (GPUs) \[63\], Field Programmable Gate
Arrays (FPGAs) \[64\], or dedicated systems-on-a-chip \[65\].
Extending generic compute with programmable accelerators
combines reconfigurability with fast and massively parallel
signal processing, and leads to platforms and systems which
are easier to upgrade, maintain, and source when compared to
dedicated application-specific integrated circuits (ASICs).

A. Implementation and Research Directions

Hardware Abstractions. To facilitate implementation, the
O-RAN Alliance Working Groups (WGs) 6 and 7 are focused
on developing a set of specifications around an Acceleration Abstraction Layer (AAL) for the RAN and white-box
hardware requirements for specific network functions \[66,\
67\]. AALs are designed to establish standardized APIs that
facilitate communication between dedicated hardware-based
processors and the O-RAN softwarized infrastructure. This
interaction covers various functions, such as channel coding/decoding and forward error correction \[66, 67\]. However,
managing and exposing GPUs, FPGAs, and CPUs or system
on chips with varying capabilities with a shared API is
challenging because of their heterogeneity. The authors of \[63\]
review the approach adopted by the O-RAN WG 6, which is
based on abstracting and providing access to functionalities
required by the 5G physical layer, but future research could
look into alternatives which are not tightly coupled to specific
physical layer features but can be extended and supported
across multiple generations.
Networking, Compute, and Energy Efficiency. The sec-

* * *

Fig. 4: RAN disaggregation and open interfaces. Adapted from \[19\].

devices, which are often highly optimized and accelerated by
dedicated ASICs. Possible approaches can rely on statistical
multiplexing and scaling, so that savings are achieved on
average compared to traditional systems, and on improvements
and optimization related to software design and hardware
platforms targeting telecom and DSP workloads.

V. OPEN RAN PRINCIPLE #1: DISAGGREGATION AND
OPEN INTERFACES

The transition to software-based platforms, where components can be implemented as software (possibly via microservices) has also ushered in more degrees of freedom in how
different layers and functionalities of the protocol stack are
mapped into atomic network functions. Traditionally, and as
discussed in Sec. II, 4G base stations are largely deployed
as monolithic units, i.e., with all the functionalities possibly
virtualized but in a single unit in proximity to the cell site.
Figure 4 illustrates how the softwarization of the 5G RAN

virtualized but in a single unit in proximity to the cell site.
Figure 4 illustrates how the softwarization of the 5G RAN
helps organizing the gNB functionalities into logical units,
the Central Unit (CU), Distributed Unit (DU), and Radio
Unit (RU). The CU is further split across the UP and CP.
The right part of Figure 4 show how different protocols in
the 3GPP stack are distributed across the units. The CU-CP
maintains the state machine of the network, keeping track on
the users state at the RRC layer. The CU-UP hosts the SDAP
and PDCP layers for managing data radio bearers. The DU
features three layers which operate in a tightly synchronized
fashion, and would thus be challenging to distribute across
non-colocated network functions—these include RLC, MAC,
and the higher part of the physical layer. The latter is split
based on operations that are carried out in the frequency
domain (scrambling, modulation, mapping to MIMO layers,
precoding, and mapping to resource elements), and in the time
domain, which are implemented in the RU. The latter also
features the Radio Frequency (RF) domain components for
signal digitalization, upconversion, and, in case, beamforming.
3GPP and O-RAN Alliance interfeces ensure that these

also supports connectivity (e.g., for handover) to other CUs,
eNBs, and to the core network components through the Xn,
X2, and NG interfaces, respectively, also defined by the 3GPP.

The DU is connected to one or multiple RUs through the
O-RAN Fronthaul interface, which builds on an enhanced
version of the common public radio interface called eCPRI.
It provides a reliable and synchronized transport layer for
control messages and user data \[69\]. The fronthaul interface
is implemented through four planes. The user plane carries in
phase and quadrature (I/Q) samples corresponding to transport blocks to be transmitted. The control plane is used to
signal when these I/Q samples need to be transmitted, and to
anticipate uplink and random access slots in which the RU
needs to process incoming signals. The synchronization plane
takes care of maintaining a nanosecond-level synchronization
between the DU and the RU through the Precision Time
Protocol (PTP). Finally, the management plane connects the
RU to the Service Management and Orchestration (SMO)
(either directly, or through the DU), and can be used to push
configurations to the radio, including, for example, beamforming codebooks \[70\].
Three other O-RAN interfaces include E2, O1, and O2.

Three other O-RAN interfaces include E2, O1, and O2.
The first connects the Near-real-time (Near-RT) RIC to the
CUs and the DUs, to retrieve telemetry and KPMs and to
enforce control or apply policies in the RAN nodes \[71\].
Data processing and control is performed by plug-and-play
components within the Near-RT RIC, i.e., the xApps. The E2
interface is designed to be flexible and extensible, with an
underlying application protocol (E2AP) to manage the connection lifecycle, and service models (E2SMs) developed on top
to provide actual functionalities. For example, E2SM KPM is
used to stream telemetry, while the E2SM RAN Control (RC)
and Cell Configuration and Control (CCC) are used for control
of UE-specific and cell-specific parameters, respectively. The
O1 interface connects the Non-real-time (Non-RT) RIC and
the SMO to all the relevant RAN functions, as shown in
Fig. 4. It is used to retrieve files, data, and configurations at a
slower time scale compared to the E2 interface, and to push
configurations and updates. It also maintains a heartbeat and
provisions new services. The O2 is the interface between the
SMO and the O-Cloud for service deployment on the cloud infrastructure. Finally, the A1 interface connects the two RICs,
with the Non-RT RIC pushing policies and, if needed, external
information elements to the Near-RT RIC and its xApps.

A. Implementation and Research Directions

Automated Interface Generation. While openness allows
for a mix and match of components and vendors, it comes with
its own challenges associated to integration, interoperability,
and interface generation. A key research direction involves
understanding how AI and natural language processing tools
can be used to automatically generate the software implementation for the open interfaces based on their specifications.
Similarly, AI can be used to automate testing and facilitate the
integration of products from different vendors, identifying the
incompatibilities, possible divergence of the implementations
from the specifications, and security issues. Automated interface generation and testing decreases the time it takes from
specification development to implementation availability, and
reduces one of the pain and risk points associated to O-RAN,
i.e., interoperability in a disaggregated environment.
Extending E2 Service Models. Similarly, automated inter-

Extending E2 Service Models. Similarly, automated interface generation can also encompass the E2 service models. As
discussed above, the O-RAN Alliance has defined a basic set
of service models for RAN control and streaming of telemetry
and KPMs. Additional service models, however, need to be
developed to enable additional functionalities and control, as,
for example, for spectrum sharing and Integrated Access and
Backhaul (IAB) optimization (as discussed in Sec. IX). A
system design challenge involves creating service models for
a wide range of functionalities and emerging use cases while
being able to operate within parameters specified by the 3GPP
for its CUs and DUs. Additionally, it is important to define
specific profiles and a fundamental set of features that must
be incorporated by RAN equipment and software vendors to
achieve O-RAN compliance.
Interface Efficiency. An additional research area is related

VI. OPEN RAN PRINCIPLE #2: TOWARD CLOUD-NATIVE
APPROACHES FOR THE RAN
Above-discussed softwarization and virtualization are posi-

Interface Efficiency. An additional research area is related
to the efficiency of the data and control communications over
the open interfaces, especially for those which are data-rate
intensive and extremely sensitive to timing, e.g., the fronthaul
interface. In this case, the data rate of the interface should not
scale linearly with the bandwidth used for wireless communications, e.g., using advanced compression and aggregation
techniques, to enable efficient massive MIMO, bandwidth scaling beyond the 400 MHz of 3GPP NR, or carrier aggregation.
Theoretical models are also needed to understand the impact
on latency and energy efficiency as compared to monolithic
architectures, considering both functionalities in 3GPP user
and control planes and control and optimization introduced
by Open RAN elements.

hardware components required to execute, optimize, and manage the RAN. Such cloud can span one or multiple locations
over cell sites, the edge, regional data centers, and beyond \[72\].
The O-RAN O-Cloud can support the deployment of all the
components of an O-RAN system (e.g., those shown in Fig. 4),
and embeds generic compute resources as well as accelerators
for DSP and hardware for AI/ML training, making it a hybrid
and heterogeneous cloud platform \[73\] designed specifically
for virtualization challenges associated to Open RAN. The
O-Cloud does not only include hardware, but also software
(e.g., hypervisors or container engines), the SMO, and its O2
interface \[74\].
Cloud-native principles transition cellular architectures into

interface \[74\].
Cloud-native principles transition cellular architectures into
fully software driven solutions (except for the RF-related
components, e.g., antennas, RF chains, and data converters).
This comes with multiple benefits, including the possibility
of automating the provisioning and management of RAN
functionalities; multi-tenant RAN environments and neutral
host solutions (e.g., as discussed in Sec. VIII and \[56\]); and
the definition of a shared abstraction over different classes of
heterogeneous hardware.

A. Orchestration and Automation

The provisioning of open interfaces and the introduction of
a unified abstraction for the hardware open new opportunities
for orchestration and automation of the whole RAN through
Continuous Integration (CI), Continuous Deployment (CD),
and Continuous Testing (CT). CI ensures that new software
patches or features are automatically integrated with the rest of
the cellular network codebase. CD automates the deployment
of these features on the cellular network infrastructure. Finally,
CT continuously evaluates the performance, security, and
compliance of the software, without the need for manual tests
which may miss important changes or bugs.
CI, CD, and CT are best practices adopted in cloud-native

which may miss important changes or bugs.
CI, CD, and CT are best practices adopted in cloud-native
environments, enabled by a variety of workflows such as
GitOps. GitOps is a state-of-the-art methodology for orchestrating CI and CD and efficiently managing infrastructure,
which can be applied in O-RAN virtualized and cloud-based
environments, as shown in \[56\]. At its core, GitOps relies
on git repositories to serve as the authoritative source of
truth, not only for the cellular network code, but also for the
infrastructure configurations (e.g., RU parameters, O-Cloud
settings, among others). This approach emphasizes the use of
declarative configuration, where the desired state of the system
is clearly defined, making it easier to understand, review,
and audit events and configurations of the cellular network.
Automated synchronization tools or controllers (e.g., ArgoCD)
continually monitor these git repositories for changes and
automatically apply them to the target environments, ensuring
that the actual system state aligns with the defined state in git.
By tracking infrastructure configurations and only allowing
automated updates from authoritative sources, GitOps enables
rapid and reliable application delivery while minimizing the
risk of configuration drift. This methodology also fosters collaboration, code review, and security practices, as changes to
infrastructure and configurations undergo the same scrutiny as
code changes. Additionally, GitOps supports observability and monitoring to uphold performance and reliability standards,
facilitates the management of multiple environments (e.g., as
in a distributed system such as the O-Cloud), and promotes
portability across different hardware implementations.

B. Security

Open RAN security is at the forefront of the discussion
around open cellular systems, as the new network interfaces,
the virtualization, softwarization, and usage of AI/ML for
control can extend the threat surface of cellular networks.
In terms of implementation challenges, the O-RAN Alliance

In terms of implementation challenges, the O-RAN Alliance
WG 11 has developed a comprehensive set of specifications
that analyze the stakeholders and threat models for Open RAN
systems \[75–77\]. These documents highlight how Open RAN
has expanded the range of stakeholders responsible for the
security of the RAN well beyond the confines of conventional
4G and 5G networks. Beyond traditional players like vendors,
operators, and system integrators, accountability is now also
required for network functions and virtualization platform
providers, third-party developers, O-Cloud service providers,
and administrators overseeing virtualized components.
As for the threat surface, documents \[77\] and \[78\] from

As for the threat surface, documents \[77\] and \[78\] from
the O-RAN Alliance and the European Commission have
identified seven distinct threat categories, encompassing a wide
array of attacks targeting different facets of the network. These
threats span from attacks on the O-RAN infrastructure itself
(e.g., the RICs), to vulnerabilities affecting the O-Cloud, opensource code, physical infrastructure, wireless functionalities,
protocol stack, and AI/ML components. Attacks listed in
the literature can compromise the availability, integrity, and
confidentiality of the network and its data \[77, 78\]. These
attacks are associated with critical assets related to interfaces,
data, and logical components, but also subpar product quality,
underdeveloped technical specifications, supply chain tampering, and infrastructure failures \[78–81\].
The inherent openness of the platform, however, also en-

C. Deployment and Research Directions

ing, and infrastructure failures \[78–81\].
The inherent openness of the platform, however, also enables operators to deploy tools and audit components for
security—a task that proves challenging in closed solutions by
vendors. The SMO assumes a crucial role in fortifying network
operations, as it comes with a global view of the RAN network
infrastructure and performance, and can run routines to spot
anomalies and weaknesses in the system \[79\]. Similarly, the
automation and CI/CD/CT discussed above enable continuous
testing and updates of the software, seamlessly deploying
security patches in a timely fashion. Finally, technical specifications have also been issued to enforce authentication and
encryption procedures, directly addressing security concerns
unique to the O-RAN architecture \[75, 76\]. As discussed
in \[82\], encryption impacts the performance of the O-RAN
interfaces, showing that it is not a concern for the non-dataintensive E2.

orchestration—enables the dynamic adjustment of compute
resources to meet user requirements, thereby limiting power
consumption to the specific network functions in use. This
concept has been discussed in previous studies \[83, 84\]. Moreover, the closed-loop control capabilities mentioned earlier,
combined with RAN virtualization, facilitate more precise and
flexible sleep cycles for base stations and RF components.
These components typically account for the majority of power
consumption in cellular networks, as highlighted in various
surveys \[85–87\]. Future research and development efforts
should focus on bridging the gap between state-of-the-art
literature on energy efficiency and approaches that leverage
and can be deployed on the open and virtualized components
of Open RAN systems, going beyond what was possible in
yesterday’s monolithic, non-programmable deployments.
An O-Cloud with Heterogeneous Compute and Devices.

yesterday’s monolithic, non-programmable deployments.
An O-Cloud with Heterogeneous Compute and Devices.
Similarly to the need for supporting and abstracting different
hardware accelerators for virtualized RAN DSP, discussed in
Sec. IV, there are also challenges when it comes to defining
an O-Cloud spanning heterogeneous hardware platforms and
components. In this sense, the O-RAN Alliance WG 7 is
outlining the specific criteria that must be met by white box
hardware to facilitate the implementation of O-RAN-compliant
equipment, specifically when considering devices with RF
capabilities. This equipment encompasses various types such
as indoor picocells, outdoor microcells, and macrocells, all
operating within sub-6 GHz and mmWave frequency ranges.
Additionally, it includes integrated access and backhaul nodes,
as well as fronthaul gateways, all part of the architectural
components depicted in Fig. 4. The specifications provide
clarity on functional parameters relevant to specific use cases
(e.g., frequency bands, bandwidth, inter-site distance, MIMO
configurations) and outline the hardware attributes (e.g., accelerators, computing capabilities, connectivity) of these nodes.
Automated Security for the RAN. While the Open RAN

erators, computing capabilities, connectivity) of these nodes.
Automated Security for the RAN. While the Open RAN
paradigm enable visibility and security best principles, it is
necessary to design, develop, deploy, and test the algorithms
and infrastructure to actually do this. A promising research
direction focus on understanding the role that AI and Machine
Learning (ML) play to secure closed-loop control and the overall Open RAN infrastructure, e.g., through automated anomaly
detection. The literature in the last decade has mostly focused
on anomaly detection for wired networks \[88\], where SDN
and packet switching architectures provide enough insights
to identify and flag anomalous traffic flows and react with
access control mechanisms. The openness and interfaces of
Open RAN systems can be leveraged to extend automated
anomaly detection in the cellular domain \[89\], thanks to the
privileged point of view that the SMO and the RICs have on
the cellular infrastructure.
Security for the AI/ML Control. AI and ML themselves,
however, are also open to vulnerabilities, which can manifest

* * *

Fig. 5: Closed-loop control in the Open RAN architecture. Based on \[9, 19\].

is necessary to research resilient AI/ML solutions for Open which represents custom plugins to be deployed alongside CUs
RAN that safeguard against vulnerabilities stemming from and DUs. The controller discussed above can be represented
data manipulation, as, for example, employing techniques by any of the these applications (or a set of applications).
such as autoencoders (as demonstrated in \[82\]), contrastive Different
learning \[90\], and contaminated best arm identification \[91\], based
among other techniques. overall point of view on the network, as well as the timescale
of the decisions. The combination of SMO and Non-RT RIC,

VII. OPEN RAN PRINCIPLE #3: AI-BASED CLOSED-LOOP
CONTROL

The programmability of the network stack, the open interfaces, and the introduction of the RICs are crucial component for the main transformation brought along by Open
RAN: data-driven closed-loop control and optimization of
the RAN functionalities. This is a fundamental step toward
enabling autonomous networks that can self-adapt to dynamic
requirements and deployment conditions. Whilst 3GPP offers
mechanisms for closed-loop control, the O-RAN Alliance was
the first to introduce a formal framework through the near-realtime and non-real-time RICs and associated interfaces.
Figure 5 illustrates how closed-loop control is enabled

Figure 5 illustrates how closed-loop control is enabled
by, and integrates into the O-RAN architecture. A specific
RAN function (e.g., the DU) exposes telemetry, analytics,
and KPMs through the O-RAN interfaces. These are received
by a controller, which performs specific tasks based on the
data it received from the network function. This task can
be some classification or regression, to infer information on
the RAN status; prediction, to anticipate dynamics in the
network function (or in general, in the RAN) behavior; or
control, leveraging reinforcement learning, optimization, or
other data-driven techniques to determine the configuration of
the network function that best matches the current status in the
RAN. For the latter, control is sent back to the RAN function,
thus actually closing the loop.
This paradigm applies to different functions and optimiza-

which represents custom plugins to be deployed alongside CUs
and DUs. The controller discussed above can be represented
by any of the these applications (or a set of applications).
Different controllers focus on different control domains,

by any of the these applications (or a set of applications).
Different controllers focus on different control domains,
based on their capabilities, access to data and KPMs, and
overall point of view on the network, as well as the timescale
of the decisions. The combination of SMO and Non-RT RIC,
together with their rApps, has a global point of view on the
O-Cloud and on the services provisioned on the network.
These components primarily focus on determining high-level
policies and managing the lifecycle of network services with a
non-real-time granularity, i.e., the control loop is closed after
more than 1 s. While there is generally one SMO instance
per network deployment, there can be multiple Near-RT RICs
(with related xApps), which are deployed at the edge of the
network and have visibility into a cluster of tens of RAN
nodes. The control loops from the Near-RT RIC operate over
the E2 interface at a timescale between 10 ms and 1 s, and
influence the radio resource management process in the RAN
with policies but also through the dynamic reconfiguration of
RAN parameters (e.g., with E2SM RC and CCC as discussed
above). Finally, dApps interface with a single RAN function at
any given time, but are envisioned to be capable of performing
control at a timescale below 10 ms (currently not considered
by the O-RAN Alliance specifications).
Closed-loop Control Use Cases. The flexibility and ca-

Closed-loop control with Open RAN primitives allows for
the fine-tuning of mobility management and performance for
specific mobile users, e.g., by adjusting handover, load balancing, multi-connectivity, access barring, and beamforming
parameters within the RAN \[93\]. In \[94\], the authors show
that tuning handover parameters with a data-driven loop that
accounts for the bespoke requirements of individual UEs
improves throughput and spectral efficiency by an average
of 50% over traditional cell-based handover heuristics. This
flexibility facilitates the optimization of the mobile experience
for single UEs, opening new use cases and possible revenue
streams for network operators.
Resource allocation is another crucial area where closed- approaches based on the point of view of individual gNBs.
Open RAN controllers can leverage the data and telemetry to
understand user requirements and evolving contexts, and map
them into effective configurations of the slicing and scheduling
policies of the network which improve resource utilization and
quality of service for users \[95\]. Researchers have explored the
application of AI/ML-based optimization in network slicing,
scheduling, and service provisioning, adapting the network to
different slices and user needs \[96\].
This is an area where AI and ML have been widely

This is an area where AI and ML have been widely
used to drive the optimization. Experiments and demonstrations on experimental platforms like Colosseum and Arena
testbeds \[97, 98\] have showcased xApps’ capabilities to intelligently control the scheduling policies of various network slices
on base stations \[9, 99\]. Different slices with specific optimization targets, such as Enhanced Mobile Broadband (eMBB),
Ultra Reliable and Low Latency Communications (URLLC),
and Machine-type Communications (MTC), can be efficiently
managed through closed-loop control mechanisms \[99–101\].
The versatility of Open RAN extends to supporting new

managed through closed-loop control mechanisms \[99–101\].
The versatility of Open RAN extends to supporting new
applications, such as vehicular communications and industrial
Internet of Things (IoT) scenarios \[102–105\]. Open RAN’s
capabilities, like dynamic control and adaptability of Massive
MIMO configurations, can enhance mobile reliability and
robustness \[104, 106, 107\]. For industrial IoT applications,
where high reliability and precise timing are crucial, Open
RAN’s closed-loop control can adapt configurations to the
evolving conditions on factory floors \[108, 109\].
Additionally, closed-loop control facilitates the optimization

evolving conditions on factory floors \[108, 109\].
Additionally, closed-loop control facilitates the optimization
of the RAN deployment itself \[51, 110, 111\]. Researchers have
proposed zero-touch orchestration frameworks, fault-tolerant
techniques, and efficient matching schemes between different
RAN components, all contributing to better resource utilization
and overall network efficiency \[51, 112–114\]. Finally, security
of the RAN can also be enhanced through closed-loop monitoring and control, as we discuss in Sec. VI-B.

Towards A Single RAN. The bulk of the 6G RAN and
architecture standardization will be carried in 3GPP. However,
whilst the O-RAN Alliance is mainly concerned with implementation and operations of networks, promising principles
ought to be natively embedded into 3GPP standards efforts.
More research is needed which maps efforts and roadmaps
such that functional standardization (3GPP) can be aligned
with operational capabilities (O-RAN Alliance).
Expansion to New Use Cases. The current use cases,

Efficient and Explainable AI/ML for RAN Control.
While O-RAN provides the basic primitives to enable closedloop control, how to achieve this in an intelligent and efficient
way is an open challenge. The use cases discussed above rely
on a mix of heuristics and AI/ML-based solutions. As industry
transitions toward intelligent control, there is a need to identify
robust, reliable, and deployable ML solutions for wireless.
In \[19\], we discuss the AI/ML workflow that O-RAN systems
support, covering the end-to-end process, from data collection
to online inference and tuning. In this context, there are still
open questions related to (i) optimal methods for training with
offline data on systems which are fundamentally dynamic and
online, but also very sensitive to performance degradation; (ii)
Deep Reinforcement Learning (DRL) solutions that generalize
well across deployments in different areas and with varying
traffic distributions, and how generative AI can be used in
the Open RAN context; and (iii) explainability solutions
for systems with dynamic control and complex input/output
relationships, among others.
Hierarchical Control. The possibilities led by two RICs

relationships, among others.
Hierarchical Control. The possibilities led by two RICs
and the dApps, and their capability to extract data from the
network operations and to manage control decisions in the
RAN, position them ideally for stemming AI/ML use cases.
As mentioned above, AI/ML can optimize slicing decisions.
Furthermore, recent studies show traffic steering and beamforming can benefit from RICs \[94, 118\] and mutually improve
learning in the hierarchical structure of the RICs by the use
of a novel ML technique called Hierarchical Reinforcement
Learning (HRL) \[119\]. Different timescales of RICs make it
challenging to have control-loops that use information from
multiple timescales. However, certain network optimizations
require fine and coarse granularity data at the same time.
HRL allows coordination of multiple timescales, yet, further
research is needed to enable information fusion and control in
a smoother way than it is today.
Conflict Mitigation. This is a critical component when

VIII. EMERGING SYSTEM REQUIREMENTS #1: AGILE
SPECTRUM AND INFRASTRUCTURE SHARING

Let us now explore two system requirements which have
been emerging recently as part of 6G design discussions.

Conflict Mitigation. This is a critical component when
considering closed-loop control in Open RAN, especially in
a hierarchical context as discussed above or with multiple
xApps or rApps targeting the same base station. Indeed, there
are instances in which different xApps may try to control the
same parameter (direct conflict), or different parameters which
have a correlated impact on the RAN (indirect conflict, e.g.,
one xApp reduces the resources associated to a slice while
another xApp hands over multiple UEs to that slice). Future
research needs to evaluate and compare different conflict
mitigation strategies, e.g., pre-deployment or post-deployment;
based on explicit declaration of control policies or implicit
reconstruction of the conflict impact or both; among other
things. In addition, there is an open discussion in terms of
which is the architectural component that is best positioned
to perform conflict mitigation, e.g., the SMO, as discussed
in \[51\], or the Near-RT RIC, or both.

* * *

The first one, discussed in this section, is on spectrum and
infrastructure sharing. The demand for faster data rates and
reduced latency in cellular networks has led to a significant
increase in network densification \[120\]. This has also given
rise to new deployment strategies, including wireless selfbackhaul solutions (which we discuss in Sec. IX), and to
an increase in the number of private operators establishing
dedicated cellular infrastructure \[121\]. Consequently, there is
a significant portion of both capital and operational expenses
faced by both public and private operators that goes towards
acquiring access to spectrum, cell site facilities (like poles
and towers), and equipment, as reported by the FCC \[122\]
and \[123\].
These increased costs can be offset through spectrum and

These increased costs can be offset through spectrum and
infrastructure sharing. The first has been recognized as an
efficient means to enhance overall spectral utilization \[124\]
and recent estimates indicate that the adoption of infrastructure and spectrum sharing techniques could potentially
yield network operational cost savings of at least 30% in
the next five years \[125\]. Infrastructure sharing is based on
the neutral host model, where infrastructure is provided by
third-party companies, leasing physical resources to multiple
operators on a shared-tenant basis \[55, 126\], reducing the
overall infrastructure costs \[127\].
As discussed in \[56\], however, RAN and spectrum sharing

As discussed in \[56\], however, RAN and spectrum sharing
are not yet suitable for widespread adoption in multi-operator
network deployments \[128\]. This is primarily due to the
absence of mechanisms that facilitate: (i) fine-grained sharing,
where multiple tenants can share compute and spectrum slices
from the same physical infrastructure; and (ii) dynamic sharing
in licensed, unlicensed, or partially licensed bands, which
allows infrastructure owners to fully harness the statistical
multiplexing of RAN and spectrum resources and adjust
infrastructure parameters to meet tenant requirements that
can change within seconds. Consider, for example, spectrum
sharing in the Citizen Broadband Radio Service (CBRS) band.
This is a partially licensed band in the U.S. with different tiers
of prioritized or general access to 150 MHz of spectrum within
3.55 GHz and 3.7 GHz, coordinated by a spectrum access
system which currently operates on timescales in the order of
minutes \[129\]. This limitation reduces system flexibility and
eventually the efficiency in terms of spectrum utilization.
The openness and programmability introduced by the Open

dynamic management of the lifecycle of network services.
This has been a challenge in traditional cellular systems,
especially when dealing with complex software services like
softwarized gNBs that need to be instantiated in a matter of
seconds \[51, 132, 133\].
Note that the O-RAN interfaces allow for better accountabil-

Note that the O-RAN interfaces allow for better accountability and visibility through fine-grained control over spectrum
sharing \[134\]. The Open RAN infrastructure also supports
SLA enforcement through dynamic, fine-grained resource allocation, driven by optimization engines \[51, 56, 135–138\].
This combination can enhance operators’ confidence in shared
systems and open up new possibilities for private and shared
deployment scenarios.

A. Research Directions

Open-RAN-driven spectrum and infrastructure sharing can
go beyond the capabilities discussed in the cognitive radio
literature \[139, 140\] by combining data- and ML-driven approaches, a hierarchical control structure with decentralized
and centralized endpoints, and automation and softwarization,
with the end goal of developing sharing approaches that
can interact with and are practically deployable in 3GPP
cellular systems. Nonetheless, lessons learned and algorithms
developed for cognitive radio systems can be analyzed and
considered within the Open RAN framework. In addition, we
identify two promising research directions as follows.
Native Sensing and Sharing. While Open RAN principles

identify two promising research directions as follows.
Native Sensing and Sharing. While Open RAN principles
enable spectrum and infrastructure sharing, how to actually
(i) sense and detect spectrum usage and allocations and
(ii) distribute resources to the various tenants are still a
challenge. Future research efforts can explore the practical
implementation of Open-RAN-driven sensing, along with both
reactive and proactive spectrum adaptation solutions, taking
into account both 3GPP and non-3GPP systems, and licensed
and unlicensed spectrum users \[141\]. Additionally, it is essential to investigate extensions to the O-RAN architecture that
are pertinent to spectrum sharing, e.g., dApps for real-time
spectrum sensing \[92\].
Spectrum Sharing in FR-3. Another area of significant in-

IX. EMERGING SYSTEM REQUIREMENTS #2:
SELF-RECONFIGURABLE WIRELESS BACKHAULING

SELF-RECONFIGURABLE WIRELESS BACKHAULING

Spectrum Sharing in FR-3. Another area of significant interest connecting 6G and intelligent spectrum sharing through
Open RAN is sharing in the centimeter band, or FR-3, as
identified by the 3GPP and recent literature \[142\]. The frequency range between 6 and 24 GHz is of significant interest
for 6G networks, as it presents a more favorable propagation
environment if compared to the lower mmWave band in FR-2,
and, at the same time, it can potentially accommodate wider
channels compared to the overly crowded sub-6 GHz range.
This spectrum, however, is currently allocated for key services
for the military, weather and Earth monitoring, and satellite
uplinks and downlinks, calling for a dynamic spectrum sharing
approach to instantiate cellular systems when possible and
avoid harmful interference to current incumbent.

* * *

Fig. 6: Proposed extension of the O-RAN architecture to managed self-backhaul scenarios based on IAB. Adapted from \[130\].

also considering the expansion to the non-terrestrial domain
and the increasing densification of terrestrial access points in
different frequency bands, including centimeter band (or FR-
3) and millimeter waves (or FR-2). Since Release 16, 3GPP
has incorporated IAB into its specifications for NR \[143\].
IAB supports the multiplexing of backhaul traffic with UE
access traffic over the same 5G NR air interface, forming a
mesh of base stations connected wirelessly, without the need
for expensive wired fiber optical connections to every access
point, as shown in the bottom part of Fig. 6. The gNBs with
a wired connection are the IAB-Donors, while the wireless
relays are the IAB-Nodes \[144\].
Compared to prior wireless mesh networks and relaying

Compared to prior wireless mesh networks and relaying
research, IAB brings opportunities related to being fully
embedded within the 3GPP stack, including the possibility
of using the same waveform and spectrum for access and
backhaul, but also challenges for scheduling and providing
reliability across both kinds of links. In this sense, despite
reaching a mature standardization stage, there are ongoing
challenges in managing, provisioning, and optimizing integrating access and backhaul. IAB provides optimization opportunities across all the layers of the 3GPP stack. At lower layers,
specialized IAB-aware scheduling techniques are necessary
to ensure fair and efficient resource allocation among UEs
(in the access) and Mobile Terminations (MTs) for IAB-
Nodes, as discussed in \[145, 146\]. Simultaneously, proper
management of temporal and spatial scheduling for IABs
flows is vital to minimize interference \[147\]. Furthermore,
adaptive topology reconfiguration mechanisms are required to
maintain resilience against link failures, traffic imbalances,
and irregular user distribution, as explored in \[148\]. These
advanced management procedures demand control primitives
beyond what the 3GPP has specified.
To this end, the Open RAN paradigm can introduce a shift
in how IAB systems are managed, through programmability,

To this end, the Open RAN paradigm can introduce a shift
in how IAB systems are managed, through programmability,
softwarization, and disaggregation. The programmatic control
of RAN components via open interfaces and centralized control loops, as described in \[19\], holds significant potential
for optimizing and managing IAB. The authors of \[130\]

discuss how the existing O-RAN architecture can evolve to
accommodate IAB control, enabling data-driven control for
IAB. The proposed architecture is illustrated in Fig. 6. The
main component is the extension of O-RAN interfaces (e.g.,
E2, O1, and O2) to the wireless domain of the network,
through dedicated tunnels which allow the RICs and SMO
to reach the DU and MT in the IAB-Nodes.

A. Research Directions

Extension of Open and Intelligent IAB to NTN. 6G
networks will likely integrate components associated to NTN,
either for backhaul or also directly for access \[41, 149\]. How
to extend O-RAN-based IAB and self-backhaul optimization
to the NTN domain is an open challenge \[150\]. Here future
research efforts intersect with the discussions on support for
NTN in the cellular architecture \[151, 152\], including (i)
evaluations of different splits for the gNBs and IAB systems
(i.e., on whether the satellite is a physical layer relay or if
it comes with the higher layers of the stack); (ii) mobility
and handover management across the terrestrial and the NTN
domain; and (iii) network topology and architectures for the
NTN component, which could be served by a variety of
non-terrestrial devices including satellites in different orbits,
Unmanned Aerial Vehicles (UAVs), baloons, among others.
In this context, the intelligent control, the openness, and the
softwarization brought by the Open RAN can enable dynamic
optimization of the resources and of the topology tree, thus
it will be important to manage the mobility and dynamics
associated to the NTN systems.
Integration of IAB Nodes, Virtualization, and Dynamic

associated to the NTN systems.
Integration of IAB Nodes, Virtualization, and Dynamic
Scaling. One of the key advantages of the softwarized infrastructure discussed in Sec. III is the possibility of seamlessly
updating, scaling, and powering on/off network functions
according to the actual network needs. Dynamic scaling is
a challenge when it comes to IAB nodes \[153\], as turning off
the radios in a self-backhauled device implies breaking the
link and communications with the upstream and downstream
nodes. The latter would also need to be reallocated to other
upstream nodes, causing multiple handovers and reconnection attempts. Therefore, there is significant interest toward design
of multi-radio IAB nodes, which exploit multi-connectivity
with low-power radios or wake-up radios to enable dynamic
scaling of resources in an IAB tree.

X. CONCLUSIONS

In this paper, we reviewed how the Open RAN paradigm
is a key enabler of innovations in 6G networks, considering a
system-level and architectural perspective. We have discussed
key requirements for 6G, including energy efficiency, coverage, resiliency, cost and complexity, and innovation, and
explored how Open RAN design principles connect to such
6G requirements. For each principle, we have presented what
are the open challenges associated to its full development and
deployment in commercial networks.
This tutorial has highlighted how openness, virtualization,

REFERENCES

\[1\]S. Parkvall, E. Dahlman, A. Furuskar, and M. Frenne, “NR: The
new 5G radio access technology,” IEEE Communications Standards
Magazine, vol. 1, no. 4, pp. 24–30, December 2017.
\[2\]Z. Pi and F. Khan, “An introduction to millimeter-wave mobile broad-

\[2\]Z. Pi and F. Khan, “An introduction to millimeter-wave mobile broadband systems,” IEEE Communications Magazine, vol. 49, no. 6, pp.
101–107, June 2011.
\[3\]S. Rangan, T. S. Rappaport, and E. Erkip, “Millimeter-wave cellular

101–107, June 2011.
\[3\]S. Rangan, T. S. Rappaport, and E. Erkip, “Millimeter-wave cellular
wireless networks: Potentials and challenges,” Proceedings of the IEEE,
vol. 102, no. 3, pp. 366–385, March 2014.
\[4\]T. L. Marzetta, “Noncooperative cellular wireless with unlimited

\[9\]L. Bonati, S. D’Oro, M. Polese, S. Basagni, and T. Melodia, “Intelligence and Learning in O-RAN for Data-driven NextG Cellular
Networks,” IEEE Communications Magazine, vol. 59, no. 10, pp. 21–
27, October 2021.
\[10\]S. Mollahasani, M. Erol-Kantarci, M. Hirab, H. Dehghan, and R. Wil-

27, October 2021.
\[10\]S. Mollahasani, M. Erol-Kantarci, M. Hirab, H. Dehghan, and R. Wilson, “Actor-critic learning based QoS-aware scheduler for reconfigurable wireless networks,” IEEE Trans. on Netw. Sci. Eng., pp. 1–10,
2021.
\[11\]P. Yang, Y. Xiao, M. Xiao, and S. Li, “6G Wireless Communications:

\[5\]S.-Y. Lien, S.-L. Shieh, Y. Huang, B. Su, Y.-L. Hsu, and H.-Y. Wei,
“5G New Radio: Waveform, Frame Structure, Multiple Access, and
Initial Access,” IEEE Communications Magazine, vol. 55, no. 6, pp.
64–71, June 2017.
\[6\]ITU-R, “IMT Vision - Framework and overall objectives of the future

\[7\]“List of 5G NR networks.” \[Online\]. Available: [https://en.wikipedia](https://en.wikipedia/).
org/wiki/List of 5G NR networks
\[8\]A. Narayanan, M. I. Rochman, A. Hassan, B. S. Firmansyah, V. Sathya,

\[13\]W. Saad, M. Bennis, and M. Chen, “A Vision of 6G Wireless Systems:
Applications, Trends, Technologies, and Open Research Problems,”
IEEE Network, vol. 34, no. 3, pp. 134–142, May 2020.
\[14\]S. Dang, O. Amin, B. Shihada, and M.-S. Alouini, “What should 6G
be?” Nature Electronics, vol. 3, no. 1, pp. 20–29, 2020.
\[15\]K. B. Letaief, W. Chen, Y. Shi, J. Zhang, and Y. A. Zhang, “The
Roadmap to 6G: AI Empowered Wireless Networks,” IEEE Communications Magazine, vol. 57, no. 8, pp. 84–90, August 2019.
\[16\]H. Tataria, M. Shafi, M. Dohler, and S. Sun, “Six Critical Challenges
for 6G Wireless Systems: A Summary and Some Solutions,” IEEE
Vehicular Technology Magazine, vol. 17, no. 1, pp. 16–26, March 2022.
\[17\]N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson,
J. Rexford, S. Shenker, and J. Turner, “OpenFlow: Enabling Innovation
in Campus Networks,” ACM SIGCOMM Computer Communication
Review, vol. 38, no. 2, pp. 69–74, March 2008.
\[18\]P. Bosshart, D. Daly, G. Gibb, M. Izzard, N. McKeown, J. Rexford,
C. Schlesinger, D. Talayco, A. Vahdat, G. Varghese, and D. Walker,
“P4: Programming Protocol-Independent Packet Processors,” SIG-
COMM Computer Communication Review, vol. 44, no. 3, p. 87–95,
July 2014.
\[19\]M. Polese, L. Bonati, S. D’Oro, S. Basagni, and T. Melodia, “Understanding Architecture, Interfaces, Algorithms, and,
Research O-RAN: Challenges,” IEEE Communications Surveys Security, & Tutorials
vol. 25, no. 2, pp. 1376–1411, Second quarter 2023.
\[20\]L. Bonati, M. Polese, S. D’Oro, S. Basagni, and T. Melodia, “Open,
programmable, and virtualized 5G networks: State-of-the-art and the
road ahead,” Computer Networks, vol. 182, pp. 1–28, December 2020.
\[21\]H. Lee, J. Cha, D. Kwon, M. Jeong, and I. Park, “Hosting AI/ML Workflows on O-RAN RIC Platform,” in Proceedings of IEEE GLOBECOM
Workshops, December 2020.
\[22\]A. S. Abdalla, P. S. Upadhyaya, V. K. Shah, and V. Marojevic, “Toward
Next Generation Open Radio Access Network–What O-RAN Can
and Cannot Do!” arXiv preprint arXiv:2111.13754 \[cs.NI\], November
2021.
\[23\]A. Garcia-Saavedra and X. Costa-Perez, “O-RAN: Disrupting the
Virtualized RAN Ecosystem,” IEEE Communications Standards Magazine, pp. 1–8, 2021.
\[24\]B. Brik, K. Boutiba, and A. Ksentini, “Deep Learning for B5G
Open Radio Access Network: Evolution, Survey, Case Studies, and
Challenges,” IEEE Open Journal of the Communications Society, pp.
1–1, January 2022.
\[25\]A. Arnaz, J. Lipman, M. Abolhasan, and M. Hiltunen, “Toward
Integrating Intelligence and Programmability in Open Radio Access
Networks: A Comprehensive Survey,” IEEE Access, vol. 10, pp.
67 747–67 770, 2022.
\[26\]D. Wypior, M. Klinkowski, and I. Michalski, “Open RAN—
Radio Access Network Evolution, Benefits and Market Trends,”
Applied Sciences, vol. 12, no. 1, 2022. \[Online\]. Available:
[https://www.mdpi.com/2076-3417/12/1/408](https://www.mdpi.com/2076-3417/12/1/408)
\[27\]3GPP, “NR; NR and NG-RAN Overall description; Stage-2,”
3rd Generation Partnership Project (3GPP), Technical Specification
(TS) 38.300, 07 2022, version 17.1.0. \[Online\]. Available: http:
// [www.3gpp.org/DynaReport/38300.htm](http://www.3gpp.org/DynaReport/38300.htm)
\[28\]——, “Network Architecture,” 3rd Generation Partnership Project
(3GPP), Technical Specification (TS) 23.002, 3 2018, version 15.0.0.
\[Online\]. Available: [http://www.3gpp.org/DynaReport/23002.htm](http://www.3gpp.org/DynaReport/23002.htm)
\[29\]M. Condoluci and T. Mahmoodi, “Softwarization and virtualization
in 5G mobile networks: Benefits, trends and challenges,” Computer
Networks, vol. 146, pp. 65–84, December 2018.
\[30\]X. Foukas, F. Sardis, F. Foster, M. K. Marina, M. A. Lema, and
M. Dohler, “Experience Building a Prototype 5G Testbed,” in Proceedings of the Workshop on Experimentation and Measurements in
5G, ser. EM-5G’18. Heraklion, Greece: Association for Computing
Machinery, 2018, p. 13–18.
\[31\]M. Shafi, A. F. Molisch, P. J. Smith, P. Z. T. Haustein, P. D.
Silva, F. Tufvesson, A. Benjebbour, and G. Wunder, “5G: A Tutorial
Overview of Standards, Trials, Challenges, Deployment, and Practice,”
IEEE Journal on Selected Areas in Communications, vol. 35, no. 6,
pp. 1201–1221, June 2017.
\[32\]M. H. Pinson, “5G Challenge Preliminary Event: Evaluating Modular,
Interoperable, Multi-Vendor, Open RAN Solutions,” Institute for
Telecommunication Sciences, Tech. Rep., 2023. \[Online\]. Available:
[https://its.ntia.gov/publications/details.aspx?pub=3319](https://its.ntia.gov/publications/details.aspx?pub=3319)
\[33\]NTIA, “Public Wireless Supply Chain Innovation Fund Homepage,”
2022\. \[Online\]. Available: [https://ntia.gov/page/innovation-fund](https://ntia.gov/page/innovation-fund)
\[34\]S. Kopsell, A. Ruzhanskiy, A. Hecker, D. Stachorra, and
N. Franchi, “Open RAN Risk Analysis,” 2021. \[Online\]. Available:

70–75, July 2019.
\[12\]M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi,
“Toward 6G networks: Use cases and technologies,” IEEE Comm.
Mag., vol. 58, no. 3, pp. 55–61, March 2020.

* * *

[https://www.bsi.bund.de/SharedDocs/Downloads/EN/BSI/Publications/](https://www.bsi.bund.de/SharedDocs/Downloads/EN/BSI/Publications/)
Studies/5G/5GRAN-Risk-Analysis.pdf? blob=publicationFile&v=7
\[35\]GSMA, “5G Energy efficiencies: Green is the new black,” Nov
2020\. \[Online\]. Available: [https://data.gsmaintelligence.com/api-web/](https://data.gsmaintelligence.com/api-web/)
v2/research-file-download?id=54165956&file=241120-5G-energy.pdf
\[36\]C. Freitag, M. Berners-Lee, K. Widdicks, B. Knowles, G. S.
Blair, and A. Friday, “The real climate and transformative
impact of ICT: A critique of estimates, trends, and regulations,”
Patterns, vol. 2, no. 9, p. 100340, 2021. \[Online\]. Available:
[https://www.sciencedirect.com/science/article/pii/S2666389921001884](https://www.sciencedirect.com/science/article/pii/S2666389921001884)
\[37\]J. Hu, Q. Wang, and K. Yang, “Energy Self-Sustainability in Full-
Spectrum 6G,” IEEE Wireless Communications, vol. 28, no. 1, pp.
104–111, February 2021.
\[38\]V. Zhang, M. Erol-Kantarci, W. Sun, Y. Dai, J. Hoydis, and M. C. Gursoy, “Guest Editorial: AI and 6G Convergence: An Energy Efficiency
Perspective,” IEEE Network, vol. 35, no. 6, pp. 10–11, November 2021.
\[39\]M. Elsayed and M. Erol-Kantarci, “AI-Enabled Future Wireless Networks: Challenges, Opportunities, and Open Issues,” IEEE Vehicular
Technology Magazine, vol. 14, no. 3, pp. 70–77, Sep. 2019.
\[40\]J. Kirby, “5G standalone networks: how vendors can
accelerate adoption,” Analysis Mason Report, 2023.
\[Online\]. Available: [https://www.analysysmason.com/contentassets/](https://www.analysysmason.com/contentassets/)
93081239f81a4ab3888b5c0d1dfaf274/analysys mason 5g sa
vendors sep2023 rma18.pdf
\[41\]M. Giordani and M. Zorzi, “Non-Terrestrial Networks in the 6G Era:
Challenges and Opportunities,” IEEE Network, vol. 35, no. 2, pp. 244–
251, March 2021.
\[42\]Y. Hokazono, H. Kohara, Y. Kishiyama, and T. Asai, “Extreme Coverage Extension in 6G: Cooperative Non-terrestrial Network Architecture
Integrating Terrestrial Networks,” in IEEE Wireless Communications
and Networking Conference (WCNC), April 2022, pp. 138–143.
\[43\]M. M. Azari, S. Solanki, S. Chatzinotas, O. Kodheli, H. Sallouha,
A. Colpaert, J. F. Mendoza Montoya, S. Pollin, A. Haqiqatnejad,
A. Mostaani, E. Lagunas, and B. Ottersten, “Evolution of Non-
Terrestrial Networks From 5G to 6G: A Survey,” IEEE Communications Surveys & Tutorials, vol. 24, no. 4, pp. 2633–2672, Fourth quarter
2022.
\[44\]J. Davies, “Network outages costing enterprise customers
millions,” 2020. \[Online\]. Available: [https://telecoms.com/504548/](https://telecoms.com/504548/)
network-outages-costing-enterprise-customers-millions/
\[45\]M. Polese, J. M. Jornet, T. Melodia, and M. Zorzi, “Toward Endto-End, Full-Stack 6G Terahertz Networks,” IEEE Communications
Magazine, vol. 58, no. 11, pp. 48–54, November 2020.
\[46\]I. F. Akyildiz, C. Han, Z. Hu, S. Nie, and J. M. Jornet, “Terahertz
band communication: An old problem revisited and research directions
for the next decade,” IEEE Transactions on Communications, vol. 70,
no. 6, pp. 4250–4285, 2022.
\[47\]T. O’Shea and J. Hoydis, “An Introduction to Deep Learning for the
Physical Layer,” IEEE Transactions on Cognitive Communications and
Networking, vol. 3, no. 4, pp. 563–575, Oct. 2017.
\[48\]T. J. O’Shea, K. Karra, and T. C. Clancy, “Learning to communicate:
Channel auto-encoders, domain specific regularizers, and attention,”
in Proc. of IEEE Intl. Symp. on Signal Processing and Information
Technology (ISSPIT), Limassol, Cyprus, Dec 2016, pp. 223–228.
\[49\]F. Dressler, “Physical Layer Resilience through Deep Learning in
Software Radios: Technical Perspective,” Communications of the ACM,
vol. 65, no. 9, pp. 82–82, 9 2022.
\[50\]F. Dressler, C. F. Chiasserini, F. H. P. Fitzek, H. Karl, R. Lo Cigno,
A. Capone, C. E. Casetti, F. Malandrino, V. Mancuso, F. Klingler,
and G. A. Rizzo, “V-Edge: Virtual Edge Computing as an Enabler
for Novel Microservices and Cooperative Computing,” IEEE Network,
vol. 36, no. 3, pp. 24–31, 5 2022.
\[51\]S. D’Oro, L. Bonati, M. Polese, and T. Melodia, “OrchestRAN:
Network Automation through Orchestrated Intelligence in the Open
RAN,” in Proceedings of IEEE INFOCOM, May 2022.
\[52\]Next G Alliance, “Next G Alliance Report:
6G Spectrum Considerations,” 2023. \[Online\]. Available: [https://www.nextgalliance.org/wp-content/uploads/dlm](https://www.nextgalliance.org/wp-content/uploads/dlm) uploads/
2023/08/6G-Spectrum-Considerations-rev1.pdf
\[53\]S. Lagen, L. Giupponi, S. Goyal, N. Patriciello, B. Bojovic, A. Demir,
and M. Beluri, “New radio beam-based access to unlicensed spectrum:
Design challenges and solutions,” IEEE Communications Surveys Tutorials, pp. 1–1, 2019.
\[54\]R. H. Tehrani, S. Vahid, D. Triantafyllopoulou, H. Lee, and K. Moessner, “Licensed spectrum sharing schemes for mobile operators: A
survey and outlook,” IEEE Communications Surveys & Tutorials,
vol. 18, no. 4, pp. 2591–2623, Fourth quarter 2016.

\[55\]J. Lahteenmaki, “The Evolution Paths of Neutral Host Businesses:
Antecedents, Strategies, and Business Models,” Telecommunications
Policy, vol. 45, no. 10, pp. 1–27, November 2021.
\[56\]L. Bonati, M. Polese, S. D’Oro, S. Basagni, and T. Melodia, “NeutRAN: An Open RAN Neutral Host Architecture for Zero-Touch RAN
and Spectrum Sharing,” IEEE Transactions on Mobile Computing, pp.
1–14, 2023.
\[57\]N. Gude, T. Koponen, J. Pettit, B. Pfaff, M. Casado, N. McKeown,
and S. Shenker, “NOX: towards an operating system for networks,”
ACM SIGCOMM computer communication review, vol. 38, no. 3, pp.
105–110, 2008.
\[58\]NextEPC. [https://nextepc.org](https://nextepc.org/). Accessed July 2020.
\[59\]Open5GS. [https://open5gs.org](https://open5gs.org/). Accessed July 2020.
\[60\]free5GC. [https://free5gc.org](https://free5gc.org/). Accessed July 2020.
\[61\]A. Checko, H. L. Christiansen, Y. Yan, L. Scolari, G. Kardaras,
M. S. Berger, and L. Dittmann, “Cloud RAN for Mobile Networks—A
Technology Overview,” IEEE Communications Surveys & Tutorials,
vol. 17, no. 1, pp. 405–426, First quarter 2015.
\[62\]xRAN Forum, “xRAN Forum Merges With C-RAN Alliance to Form
ORAN Alliance,” 2018. \[Online\]. Available: [https://www.businesswire](https://www.businesswire/).
com/news/home/20180227005673/en/
\[63\]L. Kundu, X. Lin, E. Agostini, and V. Ditya, “Hardware Acceleration
for Open Radio Access Networks: A Contemporary Overview,” arXiv
preprint arXiv:2305.09588, 2023.
\[64\]C. M. Karle, M. Kreutzer, J. Pfau, and J. Becker, “A Hardware/Software
Co-Design Approach to Prototype 6G Mobile Applications inside
the GNU Radio SDR Ecosystem Using FPGA Hardware Accelerators,” in Proceedings of the 12th International Symposium on Highly-
Efficient Accelerators and Reconfigurable Technologies, ser. HEART
’22\. Tsukuba, Japan: Association for Computing Machinery, 2022, p.
33–41.
\[65\]J. Salter, “New RISC-V hardware designs from 5G startup EdgeQ,”
2021\. \[Online\]. Available: [https://arstechnica.com/gadgets/2021/01/](https://arstechnica.com/gadgets/2021/01/)
new-risc-v-hardware-designs-from-5g-startup-edgeq/
\[66\]O-RAN Working Group 6, “O-RAN Acceleration Abstraction Layer
FEC Profiles 1.0,” O-RAN.WG6.AAL-FEC.0-v01.00 Technical Specification, July 2021.
\[67\]——, “O-RAN Acceleration Abstraction Layer General Aspects an
Principles 1.01,” O-RAN.WG6.AAL-GAnP-v01.01 Technical Specification, July 2021.
\[68\]3GPP, “NG-RAN; F1 General Aspects and Principles,” 3rd Generation
Partnership Project (3GPP), Technical Specification (TS) 38.470,
4 2022, version 17.0.0. \[Online\]. Available: [http://www.3gpp.org/](http://www.3gpp.org/)
DynaReport/38470.htm
\[69\]O-RAN Working Group 4, “O-RAN Fronthaul Control, User and
Synchronization Plane Specification 7.0,” ORAN-WG4.CUS.0-v07.00
Technical Specification, July 2021.
\[70\]——, “O-RAN Management Plane Specification 7.0,” O-
RAN.WG4.MP.0-v07.00 Technical Specification, July 2021.
\[71\]O-RAN Working Group 3, “O-RAN Near-Real-time RAN Intelligent
Controller Architecture & E2 General Aspects and Principles 2.00,”
O-RAN.WG3.E2GAP-v02.01 Technical Specification, July 2021.
\[72\]O-RAN Working Group 1, “O-RAN Architecture Description 5.00,” O-
RAN.WG1.O-RAN-Architecture-Description-v05.00 Technical Specification, July 2021.
\[73\]R. Jain and S. Paul, “Network virtualization and software defined
networking for cloud computing: a survey,” IEEE Communications
Magazine, vol. 51, no. 11, pp. 24–31, November 2013.
\[74\]O-RAN Working Group 6, “O-RAN Cloud Architecture and
Deployment Scenarios for O-RAN Virtualized RAN 2.02,” O-
RAN.WG6.CAD-v02.02 Technical Specification, July 2021.
\[75\]O-RAN Working Group 11, “O-RAN Security Requirements Specifications 4.0,” O-RAN.WG11.Security-Requirements-Specificationsv04.00 Technical Specification, October 2022.
\[76\]——, “O-RAN Security Protocols Specifications 4.0,” O-
RAN.WG11.Security-Protocols-Specifications-v04.00 Technical
Specification, July 2022.
\[77\]——, “O-RAN Security Threat Modeling and Remediation Analysis
4.0,” O-RAN.WG11.O-RAN-Threat-Model-v04.00 Technical Specification, July 2022.
\[78\]European Commission, “Cybersecurity of Open Radio Access
Networks,” May 2022. \[Online\]. Available: [https://digital-strategy.ec](https://digital-strategy.ec/).
europa.eu/en/library/cybersecurity-open-radio-access-networks
\[79\]Altiostar, “Security in Open RAN,” White paper, January 2021.
\[Online\]. Available: [http://altiostar.com/wp-content/uploads/2021/02/](http://altiostar.com/wp-content/uploads/2021/02/)
Open-RAN-Security-White-Paper-January-2021.pdf
\[80\]S. P. Jason S. Boswell, “Security considerations of Open RAN,”

* * *

Ericsson white paper, August 2021. \[Online\]. Available: [https://www](https://www/). (preprint available as arXiv:2212.11853 \[cs.NI\]), New York Area, NJ,
ericsson.com/4a67b7/assets/local/reports-papers/further-insights/doc/ USA, May 2023.
02092021-12911-security-considerations-for-cloud-ran.pdf \[100\]M. Polese, L. Bonati, S. D’Oro, S. Basagni, and T. Melodia, “ColO-
\[81\]D. Mimran, R. Bitton, Y. Kfir, E. Klevansky, O. Brodt, H. Lehmann, RAN:
Y. Elovici, and A. Shabtai, “Evaluating the Security of Open Radio Closed-loop Control on Programmable Experimental Platforms,” IEEE
Access Networks,” arXiv preprint arXiv:2201.06080, 2022. \[Online\]. Transactions on Mobile Computing, pp. 1–14, 2022.
Available: [https://arxiv.org/abs/2201.06080](https://arxiv.org/abs/2201.06080) \[101\]D. Johnson, D. Maas, and J. Van Der Merwe, “Open Source RAN Slic-
\[82\]J. Groen, S. D’Oro, U. Demir, L. Bonati, M. Polese, T. Melodia, and ing on POWDER: A Top-to-Bottom O-RAN Use Case,” in Proceedings
K. Chowdhury, “Implementing and Evaluating Security in O-RAN: of ACM MobiSys, June 2021.
Interfaces, Intelligence, and Platforms,” arXiv:2304.11125 \[cs.CR\], pp. \[102\]S. F. Abedin, A. Mahmood, N. H. Tran, Z. Han, and M. Gidlund, “Elas-
1–7, April 2023. tic O-RAN Slicing for Industrial Monitoring and Control: A Distributed
\[83\]D. Sabella, A. de Domenico, E. Katranaras, M. A. Imran, M. di Giro-Matching Game and Deep Reinforcement Learning Approach,” IEEE
lamo, U. Salim, M. Lalam, K. Samdanis, and A. Maeder, “Energy Transactions on Vehicular Technology, vol. 71, no. 10, pp. 10 808–
Efficiency Benefits of RAN-as-a-Service Concept for a Cloud-Based 10 822, Oct 2022.
5G Mobile Network Infrastructure,” IEEE Access, vol. 2, pp. 1586– \[103\]M. Hoffmann
1597, 2014. IIoT Network based on the Open RAN Architecture,” arXiv preprint
\[84\]G. Garcia-Aviles, A. Garcia-Saavedra, M. Gramaglia, X. Costa-Perez, arXiv:2302.08239, 2023.
P. Serrano, and A. Banchs, “Nuberu: Reliable RAN virtualization in \[104\]E. Moro, F. Linsalata, M. Magarini, U. Spagnolini, and A. Capone,
shared platforms,” in Proceedings of ACM MobiCom, New Orleans, “Advancing O-RAN to Facilitate Intelligence in V2X,” 2023.
LA, USA, October 2021. \[105\]I. Godor, M. Luvisotto, S. Ruffini, K. Wang, D. Patel, J. Sachs, O. Do-
\[85\]T. Pamuklu, M. Erol-Kantarci, and C. Ersoy, “Reinforcement Learning brijevic, D. P. Venmani, O. L. Moult, J. Costa-Requena, A. Poutanen,
Based Dynamic Function Splitting in Disaggregated Green Open C. Marshall, and J. Farkas, “A Look Inside 5G Standards to Support
RANs,” in Proceedings of IEEE ICC, Montreal, QC, Canada, June Time Synchronization for Smart Manufacturing,” IEEE Communica-
2021\. tions Standards Magazine, vol. 4, no. 3, pp. 14–21, Sep. 2020.
\[86\]J. Luo, Q. Chen, and L. Tang, “Reducing Power Consumption by Joint \[106\]O. T. Demir, M. Masoudi, E. Bjornson, and C. Cavdar, “Cell-Free
Sleeping Strategy and Power Control in Delay-Aware C-RAN,” IEEE Massive
Access, vol. 6, pp. 14 655–14 667, 2018. Cloud, Fronthaul, and Radio Resources,” J. Sel. Areas Commun., 2023.
\[87\]D. Lopez-Perez, A. De Domenico, N. Piovesan, G. Xinli, H. Bao, \[107\]M. Polese,
S. Qitao, and M. Debbah, “A Survey on 5G Radio Access Network “Machine
Energy Efficiency: Massive MIMO, Lean Carrier Design, Sleep Modes, Applications to 5G Cellular Networks,” IEEE Transactions on Mobile
and Machine Learning,” IEEE Communications Surveys & Tutorials, Computing, vol. 20, no. 12, pp. 3367–3382, Dec 2021.
vol. 24, no. 1, pp. 653–697, First quarter 2022. \[108\]V. Sathya, L. Zhang, and M. Yavuz, “Towards Private 5G O-RAN
\[88\]M. H. Bhuyan, D. K. Bhattacharyya, and J. K. Kalita, “Network Implementation: Performance and Business Validation,” in IEEE Future
Anomaly Detection: Methods, Systems and Tools,” IEEE Communi-Networks World Forum (FNWF), Oct 2022, pp. 676–681.
cations Surveys & Tutorials, vol. 16, no. 1, pp. 303–336, First 2014. \[109\]C. J. Tselikis, “Automated and Secure Control of Industrial Internet of
\[89\]T. Sundqvist, M. Bhuyan, and E. Elmroth, “Robust Procedural Learning Things: from WSN to Open RAN Solutions,” in International Wireless
for Anomaly Detection and Observability in 5G RAN,” IEEE Trans-Communications
actions on Network and Service Management, 2023. 556–561.
\[90\]M. Kim, J. Tack, and S. J. Hwang, “Adversarial self-supervised con-\[110\]M. Gramaglia,
trastive learning,” Advances in Neural Information Processing Systems, L. Cominardi, A. Garcia-Saavedra, and M. Fiore, “Network Intelligence
vol. 33, pp. 2983–2994, 2020. for Virtualized RAN Orchestration: The DAEMON Approach,” in Joint
\[91\]J. M. Altschuler, V.-E. Brunel, and A. Malek, “Best arm identification European Conference on Networks and Communications & 6G Summit
for contaminated bandits.” J. Mach. Learn. Res., vol. 20, no. 91, pp. (EuCNC/6G Summit), June 2022, pp. 482–487.
1–39, 2019. \[111\]D. Bega, M. Gramaglia, R. Perez, M. Fiore, A. Banchs, and X. Costa-
\[92\]S. D’Oro, M. Polese, L. Bonati, H. Cheng, and T. Melodia, “dApps: Perez, “AI-Based Autonomous Control, Management, and Orchestra-´
Distributed Applications for Real-Time Inference and Control in O-tion in 5G: From Standards to Algorithms,” IEEE Network, vol. 34,
RAN,” IEEE Communications Magazine, vol. 60, no. 11, pp. 52–58, no. 6, pp. 14–20, November 2020.
November 2022. \[112\]I. Tamim, A. Saci, M. Jammal, and A. Shami, “Downtime-Aware O-
\[93\]O-RAN Working Group 3, “O-RAN Near-Real-time RAN Intelligent RAN VNF Deployment Strategy for Optimized Self-Healing in the
Controller E2 Service Model, RAN Control 1.03,” ORAN-WG3.E2SM-O-Cloud,” arXiv preprint arXiv:2110.06060 \[cs.NI\], October 2021.
RC-v01.03 Technical Specification, October 2022. \[113\]P. Li, J. Thomas, X. Wang, A. Khalil, A. Ahmad, R. Inacio, S. Kapoor,
\[94\]A. Lacava, M. Polese, R. Sivaraj, R. Soundrarajan, B. S. Bhati, A. Parekh,
T. Singh, T. Zugno, F. Cuomo, and T. Melodia, “Programmable and ment Life-cycle of Reinforcement Learning Aided Open RAN,” arXiv
Customized Intelligence for Traffic Steering in 5G Networks Using preprint arXiv:2111.06978, 2021.
Open RAN Architectures,” IEEE Transactions on Mobile Computing, \[114\]A. Huff, M. Hiltunen, and E. P. Duarte, “RFT: Scalable and Faultpp. 1–16, 2023. Tolerant Microservices for the O-RAN Control Plane,” in Proceedings
\[95\]H. Zhang, H. Zhou, and M. Erol-Kantarci, “Federated Deep Rein-of IFIP/IEEE IM, Bordeaux, France, May 2021.
forcement Learning for Resource Allocation in O-RAN Slicing,” in \[115\]O-RAN Working Group 1, “O-RAN Use Cases Detailed Specification
IEEE Global Communications Conference (GLOBECOM), Dec 2022, 6.0,” O-RAN.WG1.Use-Cases-Detailed-Specification-v06.00 Technical
pp. 958–963. Specification, July 2021.
\[96\]M. Sharara, T. Pamuklu, S. Hoteit, V. Veque, \` and M. Erol-Kantarci, \[116\]——, “O-RAN Use Cases Analysis Report 6.0,” O-RAN.WG1.Use-
“Policy-Gradient-Based Reinforcement Learning for Computing Re-Cases-Analysis-Report-v06.00 Technical Specification, July 2021.
sources Allocation in O-RAN,” in IEEE 11th International Conference \[117\]M.
on Cloud Networking (CloudNet), Nov 2022, pp. 229–236. Networks:
\[97\]L. Bonati, P. Johari, M. Polese, S. D’Oro, S. Mohanti, M. Tehrani-Management,” IEEE Trans. Commun., vol. 66, no. 11, pp. 5621–5635,
Moayyed, D. Villa, S. Shrivastava, C. Tassie, K. Yoder, A. Bagga, Nov 2018.
P. Patel, V. Petkov, M. Seltser, F. Restuccia, A. Gosain, K. R. Chowd-\[118\]Y. Dantas,
hury, S. Basagni, and T. Melodia, “Colosseum: Large-Scale Wireless R. Gaigalas,
Experimentation Through Hardware-in-the-Loop Network Emulation,” Efficient mmWave Network Using Advantage Actor Critic Learning,”
in Proceedings of IEEE DySPAN, Virtual Conference, December 2021. Proc. of IEEE ICC, 2023.
\[98\]L. Bertizzolo, L. Bonati, E. Demirors, A. Al-Shawabka, S. D’Oro, \[119\]M. A. Habib, H. Zhou, P. E. Iturria-Rivera, M. Elsayed, M. Bavand,
F. Restuccia, and T. Melodia, “Arena: A 64-antenna SDR-based ceiling R. Gaigalas, Y. Ozcan, and M. Erol-Kantarci, “Hierarchical Reinforcegrid testing platform for sub-6 GHz 5G-and-beyond radio spectrum ment Learning Based Traffic Steering in Multi-RAT 5G Deployments,”
research,” Computer Networks, vol. 181, pp. 1–17, November 2020. Proc. of IEEE ICC, 2023.
\[99\]C. Puligheddu, J. Ashdown, C. F. Chiasserini, and F. Restuccia, “SEM-\[120\]N. Bhushan, J. Li, D. Malladi, R. Gilmore, D. Brenner, A. Damn-
O-RAN: Semantic and Flexible O-RAN Slicing for NextG Edge-janovic, R. T. Sukhavasi, C. Patel, and S. Geirhofer, “Network den-
Assisted Mobile Systems,” in Proceedings of IEEE INFOCOM 2023 sification: The dominant theme for wireless evolution into 5G,” IEEE
16

(preprint available as arXiv:2212.11853 \[cs.NI\]), New York Area, NJ,
USA, May 2023.
\[100\]M. Polese, L. Bonati, S. D’Oro, S. Basagni, and T. Melodia, “ColO-
RAN: Developing Machine Learning-based xApps for Open RAN
Closed-loop Control on Programmable Experimental Platforms,” IEEE
Transactions on Mobile Computing, pp. 1–14, 2022.
\[101\]D. Johnson, D. Maas, and J. Van Der Merwe, “Open Source RAN Slicing on POWDER: A Top-to-Bottom O-RAN Use Case,” in Proceedings
of ACM MobiSys, June 2021.
\[102\]S. F. Abedin, A. Mahmood, N. H. Tran, Z. Han, and M. Gidlund, “Elastic O-RAN Slicing for Industrial Monitoring and Control: A Distributed
Matching Game and Deep Reinforcement Learning Approach,” IEEE
Transactions on Vehicular Technology, vol. 71, no. 10, pp. 10 808–
10 822, Oct 2022.
\[103\]M. Hoffmann and P. Kryszkiewicz, “Signaling Storm Detection in
IIoT Network based on the Open RAN Architecture,” arXiv preprint
arXiv:2302.08239, 2023.
\[104\]E. Moro, F. Linsalata, M. Magarini, U. Spagnolini, and A. Capone,
“Advancing O-RAN to Facilitate Intelligence in V2X,” 2023.
\[105\]I. Godor, M. Luvisotto, S. Ruffini, K. Wang, D. Patel, J. Sachs, O. Dobrijevic, D. P. Venmani, O. L. Moult, J. Costa-Requena, A. Poutanen,
C. Marshall, and J. Farkas, “A Look Inside 5G Standards to Support
Time Synchronization for Smart Manufacturing,” IEEE Communications Standards Magazine, vol. 4, no. 3, pp. 14–21, Sep. 2020.
\[106\]O. T. Demir, M. Masoudi, E. Bjornson, and C. Cavdar, “Cell-Free
Massive MIMO in O-RAN: Energy-Aware Joint Orchestration of
Cloud, Fronthaul, and Radio Resources,” J. Sel. Areas Commun., 2023.
\[107\]M. Polese, R. Jana, V. Kounev, K. Zhang, S. Deb, and M. Zorzi,
“Machine Learning at the Edge: A Data-Driven Architecture With
Applications to 5G Cellular Networks,” IEEE Transactions on Mobile
Computing, vol. 20, no. 12, pp. 3367–3382, Dec 2021.
\[108\]V. Sathya, L. Zhang, and M. Yavuz, “Towards Private 5G O-RAN
Implementation: Performance and Business Validation,” in IEEE Future
Networks World Forum (FNWF), Oct 2022, pp. 676–681.
\[109\]C. J. Tselikis, “Automated and Secure Control of Industrial Internet of
Things: from WSN to Open RAN Solutions,” in International Wireless
Communications and Mobile Computing (IWCMC), June 2023, pp.
556–561.
\[110\]M. Gramaglia, M. Camelo, L. Fuentes, J. Ballesteros, G. Baldoni,
L. Cominardi, A. Garcia-Saavedra, and M. Fiore, “Network Intelligence
for Virtualized RAN Orchestration: The DAEMON Approach,” in Joint
European Conference on Networks and Communications & 6G Summit
(EuCNC/6G Summit), June 2022, pp. 482–487.
\[111\]D. Bega, M. Gramaglia, R. Perez, M. Fiore, A. Banchs, and X. Costa-
Perez, “AI-Based Autonomous Control, Management, and Orchestra-´
tion in 5G: From Standards to Algorithms,” IEEE Network, vol. 34,
no. 6, pp. 14–20, November 2020.
\[112\]I. Tamim, A. Saci, M. Jammal, and A. Shami, “Downtime-Aware O-
RAN VNF Deployment Strategy for Optimized Self-Healing in the
O-Cloud,” arXiv preprint arXiv:2110.06060 \[cs.NI\], October 2021.
\[113\]P. Li, J. Thomas, X. Wang, A. Khalil, A. Ahmad, R. Inacio, S. Kapoor,
A. Parekh, A. Doufexi, A. Shojaeifard et al., “RLOps: Development Life-cycle of Reinforcement Learning Aided Open RAN,” arXiv
preprint arXiv:2111.06978, 2021.
\[114\]A. Huff, M. Hiltunen, and E. P. Duarte, “RFT: Scalable and Fault-
Tolerant Microservices for the O-RAN Control Plane,” in Proceedings
of IFIP/IEEE IM, Bordeaux, France, May 2021.
\[115\]O-RAN Working Group 1, “O-RAN Use Cases Detailed Specification
6.0,” O-RAN.WG1.Use-Cases-Detailed-Specification-v06.00 Technical
Specification, July 2021.
\[116\]——, “O-RAN Use Cases Analysis Report 6.0,” O-RAN.WG1.Use-
Cases-Analysis-Report-v06.00 Technical Specification, July 2021.
\[117\]M. Chen, W. Saad, and C. Yin, “Virtual Reality Over Wireless
Networks: Quality-of-Service Model and Learning-Based Resource
Management,” IEEE Trans. Commun., vol. 66, no. 11, pp. 5621–5635,
Nov 2018.
\[118\]Y. Dantas, P. E. Iturria-Rivera, H. Zhou, M. Bavand, M. Elsayed,
R. Gaigalas, and M. Erol-Kantarci, “Beam Selection for Energy-
Efficient mmWave Network Using Advantage Actor Critic Learning,”
Proc. of IEEE ICC, 2023.
\[119\]M. A. Habib, H. Zhou, P. E. Iturria-Rivera, M. Elsayed, M. Bavand,
R. Gaigalas, Y. Ozcan, and M. Erol-Kantarci, “Hierarchical Reinforcement Learning Based Traffic Steering in Multi-RAT 5G Deployments,”
Proc. of IEEE ICC, 2023.
\[120\]N. Bhushan, J. Li, D. Malladi, R. Gilmore, D. Brenner, A. Damnjanovic, R. T. Sukhavasi, C. Patel, and S. Geirhofer, “Network densification: The dominant theme for wireless evolution into 5G,” IEEE

* * *

Communications Magazine, vol. 52, no. 2, pp. 82–89, 2014.
\[121\]M. Wen, Q. Li, K. J. Kim, D. Lopez-P ´ erez, ´ O. A. Dobre, H. V. \[142\]Next
Poor, P. Popovski, and T. A. Tsiftsis, “Private 5G networks: Concepts,
architectures, and research landscape,” IEEE Journal of Selected Topics
in Signal Processing, vol. 16, no. 1, pp. 7–25, January 2022. \[143\]T. Inoue, “5G NR Release 16 and Millimeter Wave Integrated Access
\[122\]Federal Communications Commission, “WC Docket No. 18-89,” DA
21-355, 2021. \[Online\]. Available: [https://tinyurl.com/2maycpp3](https://tinyurl.com/2maycpp3) \[144\]M. Polese, M. Giordani, T. Zugno, A. Roy, S. Goyal, D. Castor, and
\[123\]E. J. Oughton and Z. Frias, “The cost, coverage and rollout implications
of 5G infrastructure in Britain,” Telecommunications Policy, vol. 42,
no. 8, pp. 636–652, 2018.
\[124\]L. Zhang, M. Xiao, G. Wu, M. Alam, Y. Liang, and S. Li, “A survey \[145\]Y. Zhang, V. Ramamurthi, Z. Huang, and D. Ghosal, “Co-optimizing
of advanced techniques for spectrum sharing in 5G networks,” IEEE
Wireless Communications, vol. 24, no. 5, pp. 44–51, October 2017.
\[125\]C. Gabriel, “The impact of 5G and next-generation networks on
mobile OPEX spending,” Analysis Mason report, 2018. \[Online\]. \[146\]B.
Available: [https://tinyurl.com/yckndswz](https://tinyurl.com/yckndswz)
\[126\]Alpha Wireless Insights Blog, “Analysis: The Neutral Host Model
is Sprouting Wings,” 2021. \[Online\]. Available: [https://alphawireless](https://alphawireless/).
com/analysis-the-neutral-host-model-is-sprouting-wings/ \[147\]M. Yu, Y. Pi, A. Tang, and X. Wang, “Coordinated parallel resource
\[127\]K. Samdanis, X. Costa-Perez, and V. Sciancalepore, “From network
sharing to multi-tenancy: The 5G network slice broker,” IEEE Communications Magazine, vol. 54, no. 7, pp. 32–39, July 2016. \[148\]S. Ranjan, P. Chaporkar, P. Jha, and A. Karandikar, “Backhaul-Aware
\[128\]STL Partners, “Neutral Host: How Open RAN and
Neutral Host paves the way for 5G,” 2022.
\[Online\]. Available: [https://stlpartners.com/articles/telco-cloud/](https://stlpartners.com/articles/telco-cloud/) \[149\]M. Vaezi, A. Azari, S. R. Khosravirad, M. Shirvanimoghaddam, M. M.
neutral-host-how-open-ran-and-neutral-host-paves-way-5g/
\[129\]M. Palola, M. Hoyhty ¨ a, P. Aho, M. Mustonen, T. Kippola, M. Heikkil ¨ a,¨
S. Yrjol¨ a, V. Hartikainen, L. Tudose, A. Kivinen, R. Ekman, J. Hallio, ¨
J. Paavola, M. Makel ¨ ainen, and T. H ¨ anninen, “Field trial of the 3.5 GHz ¨
citizens broadband radio service governed by a spectrum access system \[150\]R. Campana, C. Amatetti, and A. Vanelli-Coralli, “O-RAN based Non-
(SAS),” in Proceedings of IEEE DySPAN, Baltimore, MD, USA, March
2017.
\[130\]E. Moro, G. Gemmi, M. Polese, L. Maccari, A. Capone, and T. Melodia, “Toward Open Integrated Access and Backhaul with O-RAN,” \[151\]Z.
in 21st Mediterranean Communication and Computer Networking
Conference (MedComNet), June 2023, pp. 61–69.
\[131\]L. Sarakis, P. Trakadas, J. Martrat, S. Prior, O. Trullols-Cruces,
E. Coronado, M. Centenaro, G. Kontopoulos, E. Atxutegi, P. Gkonis \[152\]S. Saafi, O. Vikhrova, G. Fodor, J. Hosek, and S. Andreev, “AI-Aided
et al., “Cost-efficient 5G non-public network roll-out: The Affordable5G approach,” in Proceedings of IEEE MeditCom, Athens, Greece,
September 2021.
\[132\]M. Liyanage, Q.-V. Pham, K. Dev, S. Bhattacharya, P. K. R. Mad-\[153\]G. Gemmi, M. Elkael, M. Polese, L. Maccari, H. Castel-Taleb, and
dikunta, T. R. Gadekallu, and G. Yenduri, “A survey on zero touch
network and service management (ZSM) for 5G and beyond networks,”
Journal of Network and Computer Applications, vol. 203, pp. 1–27,
July 2022.
\[133\]C. Benzaid and T. Taleb, “AI-Driven Zero Touch Network and Service
Management in 5G and Beyond: Challenges and Research Directions,”
IEEE Netw., vol. 34, no. 2, pp. 186–194, March 2020.
\[134\]GSMA, “Spectrum Sharing,” GSMA Public Policy Position, June
2021\. \[Online\]. Available: [https://tinyurl.com/5bju2k44](https://tinyurl.com/5bju2k44)
\[135\]S. K. Sharma, T. E. Bogale, L. B. Le, S. Chatzinotas, X. Wang, and
B. Ottersten, “Dynamic spectrum sharing in 5G wireless networks
with full-duplex technology: Recent advances and research challenges,”
IEEE Communications Surveys & Tutorials, vol. 20, no. 1, pp. 674–
707, February 2017.
\[136\]X. Foukas, M. K. Marina, and K. Kontovasilis, “Iris: Deep reinforcement learning driven shared spectrum access architecture for
indoor neutral-host small cells,” IEEE Journal on Selected Areas in
Communications, vol. 37, no. 8, pp. 1820–1837, August 2019.
\[137\]P. Caballero, A. Banchs, G. De Veciana, and X. Costa-Perez, ´ “Network Slicing Games: Enabling Customization in Multi-Tenant Mobile
\
Networks,” IEEE/ACM Transactions on Networking, vol. 27, no. 2, pp.
662–675, April 2019. architectures for future generations of cellular networks (5G and beyond), in
\[138\]P. Caballero, A. Banchs, G. de Veciana, and X. Costa-Perez, “Multi-´ particular for millimeter-wave and terahertz networks, spectrum sharing and
Tenant Radio Access Network Slicing: Statistical Multiplexing of passive/active user coexistence, open RAN development, and the performance
Spatial Loads,” IEEE/ACM Transactions on Networking, vol. 25, no. 5, evaluation of end-to-end, complex networks. He has contributed to O-RAN
pp. 3044–3058, October 2017. technical specifications and submitted responses to multiple FCC and NTIA
\[139\]Y.-C. Liang, K.-C. Chen, G. Y. Li, and P. Mahonen, “Cognitive radio notice
networking and communications: an overview,” IEEE Transactions on Committee on Radio Frequency Allocations of the American Meteorological
Vehicular Technology, vol. 60, no. 7, pp. 3386–3407, Sep. 2011. Society
\[140\]M. T. Masonta, M. Mzyece, and N. Ntlatlapa, “Spectrum Decision in serving as TPC co-chair for WNS3 2021-2022, as an Associate Technical
Cognitive Radio Networks: A Survey,” IEEE Communications Surveys Editor for the IEEE Communications Magazine, and has organized the Open
& Tutorials, vol. 15, no. 3, pp. 1088–1107, Third 2013. 5G Forum in Fall 2021. He is a Member of the IEEE.
\[141\]L. Baldesi, F. Restuccia, and T. Melodia, “ChARM: NextG Spectrum
Sharing Through Data-Driven Real-Time O-RAN Dynamic Control,”
in Proceedings of IEEE INFOCOM, May 2022, preprint available as

arXiv:2201.06326 \[cs.NI\].
\[142\]Next G Alliance Report, “6G Spectrum Considerations,” 2023.
\[Online\]. Available: [https://www.nextgalliance.org/wp-content/uploads/](https://www.nextgalliance.org/wp-content/uploads/)
dlm uploads/2023/08/6G-Spectrum-Considerations-rev1.pdf
\[143\]T. Inoue, “5G NR Release 16 and Millimeter Wave Integrated Access
and Backhaul,” in IEEE Radio and Wireless Symposium (RWS), 2020.
\[144\]M. Polese, M. Giordani, T. Zugno, A. Roy, S. Goyal, D. Castor, and
M. Zorzi, “Integrated Access and Backhaul in 5G mmWave Networks:
Potential and Challenges,” IEEE Communications Magazine, vol. 58,
no. 3, pp. 62–68, 2020.
\[145\]Y. Zhang, V. Ramamurthi, Z. Huang, and D. Ghosal, “Co-optimizing
performance and fairness using weighted pf scheduling and iab-aware
flow control,” in IEEE Wireless Communications and Networking
Conference (WCNC), 2020.
\[146\]B. Zhang, F. Devoti, I. Filippini, and D. De Donno, “Resource
allocation in mmWave 5G IAB networks: a reinforcement learning
approach based on column generation,” Computer Networks, vol. 196,
p. 108248, 2021.
\[147\]M. Yu, Y. Pi, A. Tang, and X. Wang, “Coordinated parallel resource
allocation for integrated access and backhaul networks,” Computer
Networks, vol. 222, p. 109533, 2023.
\[148\]S. Ranjan, P. Chaporkar, P. Jha, and A. Karandikar, “Backhaul-Aware
Cell Selection Policies in 5G IAB Networks,” in IEEE Wireless
Communications and Networking Conference (WCNC), 2021.
\[149\]M. Vaezi, A. Azari, S. R. Khosravirad, M. Shirvanimoghaddam, M. M.
Azari, D. Chasaki, and P. Popovski, “Cellular, Wide-Area, and Non-
Terrestrial IoT: A Survey on 5G Advances and the Road Toward 6G,”
IEEE Communications Surveys & Tutorials, vol. 24, no. 2, pp. 1117–
1174, 2022.
\[150\]R. Campana, C. Amatetti, and A. Vanelli-Coralli, “O-RAN based Non-
Terrestrial Networks: Trends and Challenges,” in Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G
Summit), 2023, pp. 264–269.
\[151\]Z. Abdullah, S. Kisseleff, E. Lagunas, V. N. Ha, F. Zeppenfeldt,
and S. Chatzinotas, “Integrated Access and Backhaul via Satellites,”
in IEEE PIMRC, 2023. \[Online\]. Available: [https://arxiv.org/abs/2304](https://arxiv.org/abs/2304).
01304
\[152\]S. Saafi, O. Vikhrova, G. Fodor, J. Hosek, and S. Andreev, “AI-Aided
Integrated Terrestrial and Non-Terrestrial 6G Solutions for Sustainable
Maritime Networking,” IEEE Network, vol. 36, no. 3, pp. 183–190,
May 2022.
\[153\]G. Gemmi, M. Elkael, M. Polese, L. Maccari, H. Castel-Taleb, and
T. Melodia, “Joint Routing and Energy Optimization for Integrated
Access and Backhaul with Open RAN,” in IEEE GlobeCom, 2023.
\[Online\]. Available: [https://arxiv.org/pdf/2309.05059](https://arxiv.org/pdf/2309.05059)

Networks,” IEEE/ACM Transactions on Networking, vol. 27, no. 2, pp. in the analysis and development of protocols and
architectures for future generations of cellular networks (5G and beyond), in
\[138\]P. Caballero, A. Banchs, G. de Veciana, and X. Costa-Perez, “Multi-particular for millimeter-wave and terahertz networks, spectrum sharing and
of passive/active user coexistence, open RAN development, and the performance
Spatial Loads,” IEEE/ACM Transactions on Networking, vol. 25, no. 5, evaluation of end-to-end, complex networks. He has contributed to O-RAN
technical specifications and submitted responses to multiple FCC and NTIA
\[139\]Y.-C. Liang, K.-C. Chen, G. Y. Li, and P. Mahonen, “Cognitive radio notice of inquiry and requests for comments, and is a member of the
networking and communications: an overview,” IEEE Transactions on Committee on Radio Frequency Allocations of the American Meteorological
Society (2022-2024). He was awarded with several best paper awards, is
\[140\]M. T. Masonta, M. Mzyece, and N. Ntlatlapa, “Spectrum Decision in serving as TPC co-chair for WNS3 2021-2022, as an Associate Technical
Cognitive Radio Networks: A Survey,” IEEE Communications Surveys Editor for the IEEE Communications Magazine, and has organized the Open
5G Forum in Fall 2021. He is a Member of the IEEE.
\[141\]L. Baldesi, F. Restuccia, and T. Melodia, “ChARM: NextG Spectrum

* * *

Mischa Dohler Mischa Dohler is now VP Emerging Technologies at Ericsson Inc. in Silicon Valley,
working on cutting-edge topics of 6G, Metaverse,
XR, Quantum and Blockchain. He serves on the
Technical Advisory Committee of the FCC and
on the Spectrum Advisory Board of Ofcom. He
is a Fellow of the IEEE, the Royal Academy of
Engineering, the Royal Society of Arts (RSA), the
Institution of Engineering and Technology (IET);
and a Distinguished Member of Harvard Square
Leaders Excellence. He is a serial entrepreneur with
5 companies; composer & pianist with 5 albums on Spotify/iTunes; and fluent

5 companies; composer & pianist with 5 albums on Spotify/iTunes; and fluent
in several languages. He has had ample coverage by national and international
press and media, and is featured on Amazon Prime.
He is a frequent keynote, panel and tutorial speaker, and has received

press and media, and is featured on Amazon Prime.
He is a frequent keynote, panel and tutorial speaker, and has received
numerous awards. He has pioneered several research fields, contributed to
numerous wireless broadband, IoT/M2M and cyber security standards, holds
a dozen patents, organized and chaired numerous conferences, was the
Editor-in-Chief of two journals, has more than 300 highly-cited publications,
and authored several books. He is a Top-1% Cited Innovator across all
science fields globally. He was Professor in Wireless Communications at
King’s College London and Director of the Centre for Telecommunications
Research from 2013-2021, driving cross-disciplinary research and innovation
in technology, sciences and arts. He is the Cofounder and former CTO
of the IoT-pioneering company Worldsensing; cofounder and former CTO
of the AI-driven satellite company SiriusInsight.AI, and cofounder of the
sustainability company Movingbeans. He also worked as a Senior Researcher
at Orange/France Telecom from 2005-2008.

Trans. on Network Science and Engineering, Elsevier Ad Hoc Networks, and Elsevier Nano Communication Networks. He has
been chairing conferences such as IEEE INFOCOM, ACM MobiSys, ACM
MobiHoc, IEEE VNC, IEEE GLOBECOM. He authored the textbooks Self-
Organization in Sensor and Actor Networks published by Wiley & Sons and
Vehicular Networking published by Cambridge University Press. He has been
an IEEE Distinguished Lecturer as well as an ACM Distinguished Speaker.
Dr. Dressler is an IEEE Fellow as well as an ACM Distinguished Member. He
is a member of the German National Academy of Science and Engineering
(acatech). He has been serving on the IEEE COMSOC Conference Council
and the ACM SIGMOBILE Executive Committee. His research objectives
include adaptive wireless networking (sub-6GHz, mmWave, visible light,
molecular communication) and wireless-based sensing with applications in ad
hoc and sensor networks, the Internet of Things, and Cyber-Physical Systems.

Melike Erol-Kantarci is Chief Cloud RAN AI ML
Data Scientist at Ericsson and Canada Research
Chair in AI-enabled Next-Generation Wireless Networks and Full Professor at the School of Electrical
Engineering and Computer Science at the University
of Ottawa. She is the founding director of the
Networked Systems and Communications Research
(NETCORE) laboratory. She has received numerous
awards and recognition. Dr. Erol-Kantarci is the coeditor of three books on smart grids, smart cities
and intelligent transportation. She has over 200 peer-
She has delivered 70+ keynotes, plenary talks and

Falko Dressler is full professor and Chair for
Telecommunication Networks at the School of
Electrical Engineering and Computer Science, TU
Berlin. He received his M.Sc. and Ph.D. degrees
from the Dept. of Computer Science, University
of Erlangen in 1998 and 2003, respectively. Dr.
Dressler has been associate editor-in-chief for IEEE
Trans. on Mobile Computing and Elsevier Computer
Communications as well as an editor for journals
such as IEEE/ACM Trans. on Networking, IEEE
Trans. on Network Science and Engineering, Elsevier Ad Hoc Networks, and Elsevier Nano Communication Networks. He has
been chairing conferences such as IEEE INFOCOM, ACM MobiSys, ACM

and intelligent transportation. She has over 200 peerreviewed publications. She has delivered 70+ keynotes, plenary talks and
tutorials around the globe. Dr. Erol-Kantarci is on the editorial board of
the IEEE Transactions on Communications, IEEE Transactions on Cognitive
Communications and Networking and IEEE Networking Letters. She has
acted as the general chair and technical program chair for many international
conferences and workshops. Her main research interests are AI-enabled
wireless networks, 5G and 6G wireless communications, smart grid and
Internet of Things. Dr. Erol-Kantarci is an IEEE ComSoc Distinguished
Lecturer, IEEE Senior member and ACM Senior Member.

Rittwik Jana is currently a Principal Wireless Network Engineer at Google. Prior to this he worked at
VMWare as Chief architect of RAN Intelligence and
at AT&T Labs Research. His interests span architecting the disaggregated RAN intelligent controller
in O-RAN, service composition of VNFs using
TOSCA, model driven control loop and automation
in ONAP, video streaming for cellular networks and
systems. He is a recipient of the AT&T Science
and Technology medal in 2016 for contributions to
model driven cellular network planning, the IEEE
Jack Neubauer memorial award in 2017 for systems work on full duplex
wireless, and of several best paper awards in wireless communications. He

Jack Neubauer memorial award in 2017 for systems work on full duplex
wireless, and of several best paper awards in wireless communications. He
holds 100 granted patents and has published over 150 technical papers at
IEEE and ACM conferences and journals.

* * *

Tommaso Melodia is the William Lincoln Smith Chair Professor with the Department of Electrical and Computer Engineering at Northeastern Univer- sity in Boston. He is also the Founding Director of the Institute for the Wireless Internet of Things and the Director of Research for the PAWR Project Office. He received his Ph.D. in Electrical and Computer Engineering from the Georgia Institute of Technology in 2007. He is a recipient of the National Science Foundation CAREER award. Prof. Melodia has served as Associate Editor of IEEE Transactions on Wireless Communications, IEEE Transactions on Mobile Computing, Elsevier Computer Networks, among others. He has served as Technical Program Committee Chair for IEEE Infocom 2018, General Chair for IEEE SECON 2019, ACM Nanocom 2019, and ACM WUWnet 2014. Prof. Melodia is the Director of Research for the Platforms for Advanced Wireless Research (PAWR) Project Office, a $100M public-private partnership to establish 4 city-scale platforms for wireless research to advance the US wireless ecosystem in years to come. Prof. Melodia’s research on modeling, optimization, and experimental evaluation of Internet-of-Things and wireless networked systems has been funded by the National Science Foundation, the Air Force Research Laboratory the Office of Naval Research, DARPA, and the Army Research Laboratory. Prof. Melodia is a Fellow of the IEEE and a Senior Member of the ACM.