1. Introduction
Smart
“Internet of Things” (IoT) systems refer to interconnected networks of devices, sensors, and objects that communicate and interact, collect and analyze data, make smart decisions based on the accumulated data, and take actions that improve the systems’ efficiency, productivity, and safety and security [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10]. These systems leverage the capabilities of IoT and data science technologies, such as networking connectivity, data analytics, and automation, to enable intelligent and efficient operations across various application domains. Some of the representative application domains include energy [
11,
12], transportation [
13,
14], industry [
15,
16], and healthcare [
17], among others.
The basic building blocks of these systems include the sensing, identification, and/or controlling unit consisting of sensors and actuators for perceiving their operating environments and executing the received instructions, the processing unit housing microprocessors and software applications for data processing and running computations within the network edge, a communication unit comprised of the radio access network (RAN) for connecting the different nodes to the core network and the services unit for orchestrating the other components of the system to provide the required services [
18].
The advancement of IoT technology heavily depends on its three pillars of computing, communication, and things (sensors and actuators). As evident by
Figure 1, smart IoT systems have diverse quality of service (QoS) and reliability requirements [
15,
19]. For example, healthcare wearables and energy grid smart meters are massive-connectivity-intensive. At the same time, self-driving vehicles and remote surgery are latency-intensive, while telemedicine and industrial control and monitoring systems are broadband-intensive [
20,
21,
22,
23].
In the last two decades, mobile technology has opened the wildest imaginations regarding innovation and creativity in the telecom sector by drastically changing how we communicate from the perspective of distance and geographical divide. As a result, the total user experience has been consistently and substantially improved by new technology generations launched on average every ten years [
24]. In a report published in 2022 by Dell’Oro, it was predicted that the global Radio Access Network (RAN) industry will surpass
$40 billion by 2026 and that Open-RAN deployment will represent more than 15% to 20% of RAN worldwide [
24,
25].
Traditional cellular networks such as 3G and 4G are built on monolithic, inflexible infrastructure, unlike 5G, which supports heterogeneity, allowing O-RAN to run on software-defined radio (SDR), an open loop instead of traditional analog hardware-based systems, which are closed-loop control [
26]. The practical implementation of 5G systems requires drastically reworking existing plug-and-play methodologies instead of flexible, new, and open paradigms for network management, deployment, and control [
27]. In this scenario, the cellular arena welcomes ground-breaking and cutting-edge networking solutions based on softwarization, openness, resource sharing, virtualization, and mobile edge-computing [
28,
29].
O-RAN technology aims at maximizing possible resource usage and sharing infrastructure, providing a unique market opportunity such as infrastructure leasing and connectivity-as-a-service (CaaS) technology. As a result, it is an appealing solution for network operators and infrastructure providers [
30]. The monolithic nature of RAN components, which supported a few vendors and were viewed by operators as black boxes, gave rise to drawbacks such as network node cooperation restrictions, poor reconfiguration, and vendor lock-in [
28].
Furthermore, O-RAN technology advocates for the disaggregation of the RAN into smaller modules: a radio unit (RU), distributed unit (DU), control unit (CU), and opening the interfaces between the different components to promote flexibility, interoperability, innovation, and reduce costs [
31]. The O-RAN Alliance is a group of companies formed to develop and promote the O-RAN concept. The alliance includes numerous companies, including mobile network operators, equipment vendors, and software providers. The alliance has developed a set of specifications and reference architectures for O-RAN, which are designed to ensure interoperability between different vendors’ equipment [
32]. Even though the O-RAN idea was initially conceived to support traditional mobile networks, next-generation wireless networks and smart IoT systems are also covered [
33]. For example, networks for IoT systems comprise a heterogeneous and growing number of devices, sensors, objects, and other components, which makes it challenging to manage and orchestrate the network. With the aid of O-RAN, the interface between the different network components can be standardized and opened for easier management and orchestration, leading to enhanced efficiency and cost savings. Furthermore, it can facilitate the development of open and standardized interfaces between IoT devices, gateways, and the edge and cloud to allow different vendors’ devices and gateways to work together seamlessly. Similarly, O-RAN can promote the development of enhanced edge intelligence solutions, i.e., by processing data close to where it is generated for transmission latency reduction [
33,
34,
35,
36].
In this study, we have systematically reviewed the literature by querying the reputable research databases using AND operation between keywords to have precise results such as ‘O-RAN’ and ‘IoT’, ‘O-RAN’ and ‘Transportation’, ‘O-RAN’ and ‘Industrial IoT’, ‘O-RAN’ and ‘Healthcare’, ‘O-RAN’ and ‘Energy’. Furthermore, the articles are scrutinized to ensure that the O-RAN concept supports the IoT systems conceptually or technically. Articles not satisfying the above criteria are not included in this research.
Table 1 lists the key terms and abbreviations in the paper.
Our major contributions to this paper are as follows.
We systematically review the applications of O-RAN in smart IoT systems, i.e., transportation, industry, healthcare, and energy.
We propose a generic three-dimensional problem space that considers IoT systems (e.g., transportation, industry, healthcare, energy), targets (e.g., reliable communication, real-time analytics, fault tolerance, interoperability, integration), and artificial intelligence and machine learning (AI/ML) schemes (e.g., reinforcement learning (RL), deep neural networks (DNNs).
We outline future research directions concerning robust and scalable solutions, interoperability and standardizations, privacy, and security. We also present a taxonomy to unveil the security threats to emerge from the O-RAN-assisted IoT systems and the feasible directions to address those threats.
The remainder of this paper is organized as follows.
Section 2 introduces the background of traditional radio access networks (RAN), the concept of O-RAN technology, AI/ML, and a brief literature review on O-RAN.
Section 3 explores the O-RAN in smart IoT systems based on our defined problem space.
Section 4 presents several challenges and future research directions. Finally,
Section 5 gives the final remarks.
2. Background
This section begins with a briefing of the traditional RAN, the basic concept of O-RAN technology, such as disaggregation, virtualization, RAN intelligent controller (RIC), and open interfaces. Then, this section introduces AI/ML techniques to make optimal or sub-optimal decisions or pattern recognition and classification, and a brief literature review on O-RAN.
2.1. Traditional Radio Access Network (RAN)
The RAN, popularly known as the access network, is the radio component of the cellular network. It is an essential element of a wireless telecommunication system that employs radio links to connect each device to other network sections. The RAN connects user equipment (UEs) over a fiber or wireless backhaul connection, such as a phone, computer, or any remotely operated machine. From the first generation (1G) of cellular networking to the fifth generation (5G), RANs have undergone several developments. The core network and radio access network underwent a significant change due to the introduction of long-term evolution (LTE) RAN by the Third Generation Partnership Project (3GPP) in the 2000s with the advent of fourth-generation (4G) technology. With 4G, system connectivity was replaced by circuit-based networks for the first time and was based on the internet protocol (IP). The first cellular networks were developed in the early years of the 2000s. Text messages, voice calls, streaming video, and audio have all become part of RAN’s capabilities. Many more types of UEs use these networks, including autonomous cars, drones, and IoT devices.
Despite these remarkable advancements, traditional RANs are monolithic nowadays (i.e., inflexible, protocol-based, proprietary-driven), and inefficient in handling a number of vast dynamic requirements brought by 5G and future 6G. This calls for open, flexible, intelligent, software-driven connectivity solutions capable of handling these requirements and resolving the traditional RAN’s deficiencies in a more cost-effective and energy-efficient manner. To address the shortcomings posed by RAN architecture, a number of research and standardization initiatives proposed O-RAN as the new paradigm shift for the future of RAN. Notwithstanding, O-RAN implementations use software-based, virtualized, and disaggregated components that are interoperable across multiple vendors and connected by open, standardized interfaces [
37,
38].
2.2. Concept of O-RAN Technology
The new O-RAN paradigm approach is software-based and adaptable, enabling data-informed smart control loops for cellular systems and openness, virtualization, and programmability of RAN functionality and components. As a result, network engineers can support new customized services on common physical infrastructures and vigorously rearrange them in response to user demand and network conditions thanks to the Open RAN. In addition, the network’s operating expenses could decrease due to greater efficiency [
26].
Disaggregation: It entails breaking down the RAN components into smaller chunks of functional modules that different vendors can implement. The following units emerged by disaggregating the O-RAN components: the baseband unit (BBU), i.e., the control unit (CU), which is responsible for control plane functions, such as radio resource management and mobility management; the distributed unit (DU), which is responsible for radio frequency (RF) processing and data plane functions, such as user data forwarding; and finally, the radio unit (RU), which is responsible for radio frequency (RF) signal reception and transmission.
Virtualization: It encompasses separating the RAN software components from the underlying hardware to allow greater flexibility and agility in the RAN and the potential for cost savings. Since software components are tightly coupled to the underlying hardware in a traditional RAN, virtualization decouples the software components from the underlying hardware to optimize the RAN’s performance for different application environments or even gives room for creating a responsive cloud-based RAN that scales or shrinks based on the available demand to ease the management of the RAN.
RAN intelligent controller (RIC): It is a key component of the O-RAN architecture. It carries out controlling and managing the RAN, and it uses software-based intelligence to optimize the performance of the network. The RIC is divided into the near-real-time RIC (NRT-RIC) and the non-real-time RIC (N-NRT-RIC). The N-NRT-RIC operates on a timescale greater than 1 s to handle non-real-time events and their respective control loops, such as configuration changes and performance optimization. The NRT-RIC operates between the 10 ms and 1 s timescale to handle near real-time events and their respective control loops, such as handovers and admission control.
D’Oro et al. [
39] proposed the concept of dApps operating in the CU and DU to extend the RIC capabilities towards handling real-time control loops such as scheduling and beamforming working on timescales of less than 10 ms, even though efforts extending the RIC concept to the RU are yet to be recorded at the time of this writing. For instance, Ko et al. [
40] proposed edgeRic to confirm that decoupling the RIC from the RAN stack and co-locating it close to the DU can support control loops operating within sub milliseconds and even microseconds. These proposed contributions extended the RIC capabilities toward addressing the latency requirements of the next-generation smart IoT systems.
Open interface: It involves the collection of standardized interfaces that enable the different components of the RAN to communicate with each other. The interfaces are designed to be open and interoperable to facilitate the seamless cooperation of different vendors.
Some of the interfaces that are defined and standardized by the O-RAN Alliance include the following: The
O1 interface, which connects the SMO with the remaining components (CU, DU, RU, and near-RT RIC) of the RAN stack to manage and orchestrate the RAN as well as enable the integration of the third-party apps. The
A1 interface, which provides a connection between the non-RT-RIC and the N-RT-RIC to send intelligent policies and telemetries to the N-RT-RIC and allow the N-RT-RIC to control the RAN intelligently. The
E2 interface, which connects the N-RT-RIC with the E2 nodes, enabling different RAN optimization and automation services like RAN control, monitoring, and configuration. Other interfaces include the
fronthaul, which connects the RU with the DU.
Midhaul enables connections between the DU and the CU, and backhaul connects the RAN with the core network.
Figure 2 describes the architecture of O-RAN as defined by O-RAN Alliance [
32]. The alliance is a global industry consortium that promotes developing and adopting open and intelligent RAN technologies. It was initiated to transform traditional RAN infrastructure by enabling greater interoperability, flexibility, and innovation in the mobile telecommunications industry. The alliance plays a pivotal role in driving innovation and openness in RAN technology, which is essential for the evolution of 5G networks and the deployment of future wireless communication technologies.
2.3. Artificial Intelligence and Machine Learning (AI/ML)
Simply put, AI is the art of enabling machines with human-like thought, reasoning, and decision-making power. It is the grand quest to breathe life into algorithms, allowing them to perceive the world, learn from experience to build a model that can reflect the target system, adapt to changing circumstances, and provide solutions to complex problems [
7,
41,
42]. The ML, which is a subset of AI, learns from the given data about the past to make better futuristic predictions. It is the digital detective, discovering patterns and insights that elude human senses in pursuing more intelligent and autonomous machines.
In this context, the typical applications of AI/ML methods include dynamic control decision-making and pattern recognition/classification techniques as described below.
Reinforcement learning (RL) is a branch of the dynamic ML paradigm where software agents learn to make decisions in an environment by receiving feedback as rewards [
43]. Through a cycle of observations, actions, and learning, the agents gradually improve their decision-making abilities to achieve long-term goals. Several RL types range from value-based, policy-based, deep RL, or actor–critic learning, which combines the two for optimized decision making in dynamic environments.
RL is applied in the O-RAN context to optimize the management and performance of the network. For example, RL dynamically allocates resources (bandwidth, power, frequencies, etc.) to different IoT devices or network slices based on real-time conditions and demands. Similarly, RL can control power consumption by optimizing transmission power levels and sleep modes of devices while ensuring connectivity and QoS. Again, RL agents can be trained to detect network anomalies and failures and then decide to implement self-healing mechanisms. This proactive approach ensures network resilience and reduces downtime. Despite RL’s promising capabilities for optimizing O-RAN and the emerging O-RAN-based IoT systems, challenges related to training data, model robustness, and real-time adaptation, among others, remain open for the research community to solve [
43,
44].
Deep neural networks (DNNs): This refers to the stacked neural network architectures that recognize patterns and extract information from structured or unstructured data. They are well suited for tasks involving images, sequences, and graph data [
41]. Convolutional neural networks (CNNs), as a good example, are applied to tasks involving grid-like data, such as images and videos. They are proficient at capturing spatial relationships within data. Recurrent neural networks (RNNs) on the other hand, are designed for handling sequential data. The architecture has connections that allow information to persist over time, which makes it exemplary at capturing temporal relationships within the data. Graph neural networks (GNNs) are essential for data representation involving graphs, i.e., data consisting of several nodes and edges. They are mostly used in applications with essential relationships between data points, such as network traffic analysis [
45].
In this context, these techniques are applied to learn the latent characteristics of the network and its users, which will guide the effective management of the network in several ways. For example, these techniques can learn the operating pattern of the network from its data to predict traffic and performance degradation, detect anomalies and intrusions, and optimize energy efficiency. However, despite DNN’s potential, striking a balance between their complexity and the complexity of the emerging O-RAN-based IoT system to satisfy the real-time performance requirement of their operating environment remains open for further research [
41,
46,
47].
2.4. Brief Review on O-RAN
A number of recent research works have shown that O-RAN can support the next-generation smart IoT systems in several ways. For example, in two separate works, researchers in Pham et al. [
48] and Pham et al. [
49] considered the Internet of drones by integrating UAVs with O-RAN architecture to optimize user routes to the core network to improve resource allocation and enhance offloading tasks. Similarly, Wang et al. [
36] proposed a computation offloading strategy that minimizes energy consumption and reduces the communication latency of O-RAN-based IoT systems. Riccio et al. [
50] leveraged the RAN intelligent controller (RIC) to enhance IoT handover management. Lini et al. [
51] emphasized that merging AI with 5G is essential in facilitating open networking (vendor independence) as well as realizing the next-generation smart IoT systems. Firouzi et al. [
35] adopted O-RAN to optimize federated learning (FL) operation in supporting distributed edge intelligence of 5G-based IoT systems. Liu et al. [
52] addressed network slicing security in cyber–physical systems (CPS), while Kougioumzidis et al. [
53] considered extending O-RAN capabilities to virtual reality applications. Vila et al. [
54] presented the network digital twins (DT) concept to reflect the RAN’s operation for taking the appropriate action. Likewise, Masaracchia et al. [
55] demonstrated the integration of DT with O-RAN as inevitable for yielding the 6G RAN.
Similarly, there are several surveys and reviews on O-RAN, its application areas, and areas of challenges. For example, Polese et al. [
31] presented a comprehensive tutorial on O-RAN, discussing its architecture, interfaces, and workflows, giving a big picture of how O-RAN aims to transform the next-generation cellular networks. In addition to covering the areas of research successes and challenges, Liyanage et al. [
56] presented both threats and possible solutions that are associated with O-RAN’s security and privacy concerns. Bitton et al. [
57] presented a systematic adversarial machine learning (AML) threat analysis for O-RAN, discussing some of the possible AML countermeasures and a method for conducting risk assessments. Abdalla et al. [
58] reviewed O-RAN capabilities. Likewise, Wu et al. [
59] reviewed network slicing management in Industrial IoT. Thus, there is room for more effort within the public domain to provide a holistic investigation of the applications of O-RAN in smart IoT systems.
3. O-RAN in Smart IoT Systems
This section explores the smart IoT systems and the O-RAN contribution in each selected domain. In particular, we first describe the problem system and then present the O-RAN application to some representative IoT systems.
3.1. Problem Space
We propose a three-dimensional problem space, as detailed by
Figure 3, where the
X-axis indicates the IoT systems (e.g., STS, SMS, others), the
Y-axis displays the targets (e.g., reliable communication, real-time analytics, fault tolerance, interoperability and integration), and the
Z-axis covers the AI/ML techniques (i.e., RL, DNN).
As shown in
Figure 3, four targets were adopted to represent the research objectives achieved by O-RAN while supporting the smart IoT systems. The targets derived from the adopted research methodology and the diligent consideration of the objectives achieved by O-RAN in these domains. The dimension of targets (
Y) are detailed below:
Reliable communication (): It entails the achievement of reliable, optimized, and secure communication infrastructure for seamless connectivity and data exchange within and between these systems by addressing challenges related to latency, bandwidth, privacy, authentication protocols, interference, etc. For example, V2X communication represents optimal transmission latency [
60], IoE beamforming represents optimized interference in the transmission medium [
61], and telemedicine represents optimum end-to-end healthcare service delivery [
62].
Real-time analytics (): It involves improving data collection, analysis, decision-making, and control by integrating edge computing, distributed analytics, and machine learning techniques to support extracting valuable insights and real-time monitoring. It also enables predictive maintenance, personalized services, etc. For example, resource allocation represents the dynamic assignment of network resources in real time [
34], and edge intelligence represents data analytics close to where it is generated for improved performance [
63,
64].
Fault tolerance (): It implies using the O-RAN concept to enhance IoT systems’ resilience and self-healing capabilities by improving the robustness of fault detection and recovery mechanisms to ensure uninterrupted operation, quick response to disruptions and security threats to minimize downtime. Some of the examples are signaling storm detection [
65] and Industrial IoT data security [
66].
Interoperability and integration (): It entails using the O-RAN concept to promote interoperability and integration among IoT devices, systems, and platforms using the standardized interfaces to enable seamless connectivity, data exchange, and integration of diverse components for efficient operations. For example, O-RAN can be used to facilitate interoperability between two or more IoT domains [
59].
It is worth noting that to make O-RAN a viable wireless network infrastructure to support IoT applications, it is critical to perform quantitative analysis with performance metrics (latency, bandwidth, scalability, capacity, coverage, range, etc.) for IoT applications based on O-RAN. For example, related to smart manufacturing systems, to carry out monitoring capability of industrial systems, the latency range must be below 100 ms, bandwidth must be in the range of 0.1–0.5 Mp/s, and reliability needs to be larger than 99.9% [
67]. In our prior study [
15], we emphasized the importance of understanding the key parameters and performance requirements of smart manufacturing systems, including types of manufacturing applications, key parameters (network size, etc.) and performance requirements (latency, reliability, etc.), discussed the challenges and limitations of existing static networking infrastructure to support applications in smart manufacturing systems with diversified performance requirements, and called for the necessity of developing dynamic and flexible/re-configurable networking infrastructures (e.g., software-defined networking) to support IoT applications with different performance requirements. Additionally, for energy-based IoT systems such as smart grids [
68], it is critical to understand the performance requirements of smart grid applications in both the transmission and distribution side based on the standardization and carry out performance evaluation of different networking architectures and protocols. As the O-RAN is a new network infrastructure, the standardization and research community shall have joint efforts on the performance gap of O-RAN to support IoT applications via modeling, simulation, and testbed development.
3.2. Smart Transportation Systems
The smart transportation system leverages advanced information communication technologies (ICT) to enhance transportation systems’ safety, efficiency, reliability, and sustainability. It integrates sensors, communication systems, data analytics, and control systems to collect and process data, providing real-time decisions, and optimize transportation operations. Nowadays, the smart transportation system is used to support the automatic control of vehicles, monitor traffic conditions, handle traffic flow, provide real-time information updates to drivers, reduce environmental pollution, and even improve air quality. With the advent of 5G, the next-generation smart transportation system aims to improve road safety, traffic efficiency, environmental sustainability, and user experience through enhanced communication and data-driven technologies [
69]. Just like the
“ultra-reliable and low-latency communication (URLLC)” capabilities to support real-time applications like autonomous driving, improving the reliability, range, and capacity of V2X communication promotes better coordination among road users [
70]. The
“enhanced mobile broadband (emBB)” applications like in-vehicle entertainment and infotainments require high-quality multimedia streaming, augmented reality navigation, and other entertainment services to passengers in moving vehicles [
71]. The
“massive machine type communication (mMTC)” enables more effective traffic management and congestion reduction strategies using dynamic traffic signal control, adaptive routing, and predictive traffic flow analysis, etc. [
72,
73].
At the same time, O-RAN could build on that to promote automation, interoperability, and cost-efficiency in the next-generation ITS by enhancing vehicular communication optimization, traffic management and control, cooperative autonomous driving, etc. For example, Huang et al. [
63] presented an improved edge intelligence supporting IoV networks. Hammami et al. [
74] leveraged the O-RAN concept to implement a centralized training but distributed execution (CTDE) strategy to enable cooperation among diverging agents operating in vehicular networks. In this manner, they addressed the resource allocation problem of V2V and V2I while satisfying their reliability and QoS requirements. Likewise, Ndikumana et al. [
64] used the NRT-RIC to create a collaboration area in a network of multiple radios and edges. Their approach can balance computation and communication, translating to a latency reduction in autonomous vehicles network [
75].
Abolhasan et al. [
76] utilized the RIC to control the proposed fuzzy-based routing protocol. The RIC handles the dynamic network topology and arrives at the optimum traffic route by generating and maintaining the link state database, which guides the multi-hop peer-to-peer (P2P) communication. Linsalata et al. [
60] presented O-RAN as a promising enabler of the next-generation vehicular communication orchestration, together with open research considerations militating against integrating the two technologies. V2X resource allocation, beam selection, relay assignment, and network DTs are among the feasible research directions in their work.
Takeaways from smart transportation systems: Supported by
Table 2, the reviewed efforts in smart transportation covered
and
, while
and
are less explored. However, despite O-RAN’s achievements in smart transportation, there is a need for smart-transportation-compatible specifications, standards, and tools facilitating the development of robust, secure, energy-efficient, adaptive solutions that operate in real life.
3.3. Smart Manufacturing Systems
Smart manufacturing applies cutting-edge technology that uses Internet-connected equipment to increase production efficiency and track processes by employing computer controls, modeling, big data, and other automation [
15]. The Industrial Internet of Things (IIoT) has specific applications, one of which is smart manufacturing. The IIoT aims to increase the reachability, effectiveness, and decision making in industrial control systems (ICS). It also offers intelligent functions, including energy consumption monitoring, environmental release management, and general safety and security for the ICS network. ICSs are hardware and software-based systems that monitor and manage industrial activities. The networking hardware and protocols that serve the ICS are known as operational technologies (OT). Supervisory control and data acquisition (SCADA), plant distribution control systems (DCSs), as well as programmable logic controllers (PLCs) are some of the common examples of OT employed in ICS [
77]. This entails integrating inter-connected IoT devices, technologies, and principles in industrial settings to improve operational efficiency, optimize processes, and enhance productivity. The smart manufacturing system leverages sensors, devices, and intelligent systems to connect and collect data from various industrial assets, machines, and equipment, enabling real-time monitoring, control of environmental activities, analysis, and automation.
From the 5G era and beyond, the smart manufacturing system targets improving seamless connectivity, ultra-reliable communication, high data rate, low latency, efficient data processing, and optimal resource allocation for various industrial applications, for example, industrial automation and remote robotics, where minimal transmission latency is the principal objective, video surveillance for real-time monitoring of industrial assets where high-speed data rates are targeted, and industrial assets tracking, where massive machine connectivity is preferred, among others [
78,
79,
80].
O-RAN will build on 5G’s intervention to yield open, scalable, resilient, cost-effective solutions that lead to a more efficient, productive, resilient, and sustainable industrial setting. This can be achieved in several ways. For example, Rahman et al. [
67] conducted a
“SWOT Analysis” to determine O-RAN’s areas of strengths that could be leveraged to support the next-generation smart manufacturing applications in satisfying their seamless connectivity requirements. Lin et al. [
81] proposed
“zero-touch” architecture to connect the diverse components of a smart factory together. The RIC was used to control the connectivity using AI/ML in a distributed passion.
Abedin et al. [
34] proposed a flexible and scalable method of slicing the IIoT network in RIC while considering the diverse QoS requirements and the available resources within the network. Their approach reduces the problem’s complexity by employing distributed game theory to match each slice to its corresponding cell base stations and then applying actor–critic learning to achieve an optimum resource allocation that facilitates IIoT monitoring and control.
Hoffmann et al. [
65] proposed an xApp that examines the control plane messages to detect malicious threats targeting IIoT devices. Tselikis et al. [
66] presented IIoT data security as critical for effectively monitoring and controlling the ecosystem, thereby examining O-RAN’s readiness towards ensuring the privacy and security of the data.
Takeaways from smart manufacturing systems: As supported by
Table 3, the reviewed efforts in smart manufacturing systems covered
,
, and
while
still needs further exploration. However, despite O-RAN’s achievements in smart manufacturing, there is a need for more efforts on SMS-compatible specifications, standards, and development, as well as testing tools facilitating the development of scalable, robust, adaptive solutions that keep the end-to-end security and privacy of the operating environment and its data in mind.
3.4. Smart Healthcare Systems
The smart healthcare system entails leveraging technological advancements in AI/ML and IoT to improve healthcare delivery, enhance medical services, improve patient outcomes, and streamline healthcare processes. The fundamental objective of smart healthcare is to provide personalized, efficient, and effective healthcare services while minimizing costs and enhancing the general quality of caregiving.
Some of the critical components of smart healthcare include IoT devices (sensors/actuators) and healthcare wearables for collecting patients’ data remotely. The data are analyzed using AI/ML models to predict the required action, processed and stored with the aid of computing infrastructures like the network edge, cloud, data centers, and other networking equipment, and supported with a user interface to guide user interaction with the system. Due to the critical nature of the system and the data in circulation, the system is supported with authentication, encryption, access control, and intrusion detection mechanisms to ensure the security and privacy of the system and data [
82].
With the benefits realized from the advancements in cellular networks like the 5G/6G, O-RAN can improve the services rendered by IHS by providing a flexible, scalable, and interoperable architecture that can support many devices, users, and systems to operate more cost-effectively and energy-efficiently.
For example, De et al. [
62] presented “OpenCare5G”, a project aimed at demonstrating 5G-enabled telemedicine, i.e., remote digital healthcare examinations in real-time. The project employs the O-RAN concept to disaggregate the network components, giving room for customization and updates and making the solution flexible and cost-effective. Likewise, Trifonov et al. [
83] demonstrated how O-RAN uses AI/ML to improve network performance and supported that a policy-driven use case could control the inactivity timer prediction of narrow-band medical IoT devices.
Takeaways from smart healthcare systems: Supported by
Table 4, the reviewed efforts in smart healthcare covered
and
, while
and
are yet to be explored. However, despite O-RAN’s achievements in smart healthcare, there is a need for more efforts to transition the several objectives in smart healthcare to reality as well as healthcare-compatible specifications, standards and tools facilitating the development of enhanced solutions that preserve the end-to-end privacy and security of the patient’s data.
3.5. Smart Energy Systems
The smart energy system, such as the smart grid, involves integrating modern communication and information technologies with the energy sector to create a more interconnected, efficient, and sustainable energy ecosystem by applying digital solutions to energy generation, distribution, consumption, and management. The smart energy system leverages data analytics, automation, and real-time communication to optimize energy systems, enhance grid stability, and support the integration of distributed energy resources (DER). Some candidate systems in smart energy systems include smart grids, energy management and optimization systems, renewable energy integration mechanisms, demand response programs, microgrids, and energy trading schemes.
With 5G and beyond 5G (B5G) networks, the smart energy system targets the revolution of the entire energy sector by supporting efficient, flexible, and responsive energy systems. This can be achieved in several ways. Starting from ensuring low latency, facilitating real-time communication between various energy assets, sensors, and control systems to maintain grid stability, respond quickly to supply and demand fluctuations, and ensure the reliability of energy services to providing the massive connectivity required by the smart energy system to connect a wide range of devices, like the smart meters, sensors, DERs, energy storage systems, and electric vehicles as well as providing the necessary bandwidth required for transmitting large amounts of data in real-time, especially for the essential applications requiring high-resolution energy consumption monitoring, real-time analytics, and video surveillance of energy infrastructures [
84]. When the O-RAN concept is extended to smart energy applications, it will yield more flexibly connected, efficient, and collaborative energy systems supporting enhanced communication, improved management of DERs, better grid stability, and increased flexibility in adapting to evolving energy demands.
For example, Mongay et al. [
61] presented 5G as a promising enabler of smart energy systems and supported that with an O-RAN-based beamforming case study. Similarly, Kundacina et al. [
85] proposed how 5G will support the next generation’s smart distributed state estimation of PMU-based WAMS in near-real-time.
Takeaways from smart energy systems: Supported by
Table 5, the reviewed efforts in smart energy systems covered
and
, while
and
remain open for further investigation. However, despite O-RAN’s achievements in smart energy systems, there is a need for more real-time operating ideas to maintain the integrity of the operating environment’s data, minimizing interference in the transmission medium, which will improve the system’s resilience to threats.
Finally, as depicted in
Table 6, the O-RAN-IoT convergence helps to minimize IoT devices’ energy consumption. In particular, computation tasks can be offloaded from the participating nodes to the network edge in the representative CPS (smart transportation, smart manufacturing, smart energy grids, smart healthcare, etc). The network edge can leverage the O-RAN to promote openness and intelligence in addition to virtualization, scalability, and flexibility provided by the proprietary C-RAN and V-RAN. The openness in the edge offers the opportunity for heterogeneous computation and communication tasks to be jointly optimized and executed for latency reduction, which translates to the cost-effective energy consumption of participating IoT devices. Taking the computation offloading case [
86] as an example, offloading points, processing speed, offloading ratio, and transmission power were jointly optimized within the edge. Since computation is carried out in the O-RAN-based edge and only the results are sent back to the participating nodes (IoT devices) via the downlink channel; this avoids traffic congestion, drastically reducing the energy consumption of IoT devices. Illustratively, in the smart transportation system, by minimizing the offloading delay in an open collaborative edge [
64], the convergence leads to a massive communication latency reduction, a stringent requirement for the effective communication of autonomous vehicles. Likewise, in the smart healthcare system, the convergence can considerably reduce the cost associated with remote healthcare giving, just like the case of the OpenCare5G project [
62], where the O-RAN concept was used to realize an interoperable architecture housing hardware and software, virtual machines and cloud-based open interfaces developed by different healthcare vendors.
5. Final Remarks
IoT systems consist of physical devices and various objects connected and integrated with monitoring and control capabilities, which can be applied to numerous critical infrastructure systems in energy, transportation, and cities, among others. With the advancement of information communication technology, IoT systems can be enabled by sensors/actuators and other enabling technologies, which support data collection and exchange and derive intelligent decisions to guide their operations based on the collected data via viable data science and machine learning technologies. The advancements in cellular networks, i.e., 5G and beyond, have enabled these systems to achieve their required connectivity. However, their reliability and QoS across various pervasive domains require more effective, economical, and adequate solutions.
In this paper, we have investigated the applications of O-RAN in this context by conducting a literature survey to determine how openness in the RAN, as a critical networking infrastructure, can support the next-generation IoT systems. This research proposes a problem space covering IoT systems: transportation, industry, healthcare, and energy; targets; reliable communication, real-time analytics, fault tolerance, interoperability and integration, and AI/ML; and RL and DNNs. We have outlined several challenges militating against integrating these concepts and future research directions guiding how to design O-RAN for IoT applications and deal with emerging security and privacy issues, natural AI solutions in O-RAN, and finally, the standardization, research, and development efforts required in the forthcoming years. In the future, we plan to extend this research by investigating the feasible simulation tools, experimental testbeds, and other supporting technologies that could be leveraged to transform the interoperability and integration () vision into fruition, as well as the systematical exploration of threats and the design of countermeasures to deal with the security and privacy () of the emerging O-RAN-based IoT system.