Dedicated Communication for Sensing: A New Use Case for Private 5G and Open RAN for mMTC

1. Introduction

IN fifth generation (5G) mobile communication, ubiquitous infrastructure and the industrial Internet of Things (IIoT) have led to the introduction of massive machine-type communication (mMTC) [1]. As mMTC applications are widely used in various industries, the service coverage of mobile networks is expanding from humans to machines. In a typical service scenario, large-scale smart factories comprise of massive sensor networks and mobile robots (Fig. 1). Sensing plays a significant role in mMTC networks and is expected to further expand its impact in the era of beyond 5G and 6G. In this context, we can consider integrated sensing and communications (ISAC) for mMTC networks, which is a key candidate technology for next-generation mobile communications [2].

Fig. 1.

An example of an mMTC network involving sensor networks and mobile robots.

Traditionally, the definition of ISAC is a system where radar sensing and wireless communication are integrated into one infrastructure [3]. As sensing and communication are both utilizing mmWave bands and employing multiple-input multipleoutput (MIMO) technology, the hardware architecture for radar sensing and mobile communication has become strikingly similar [4]. Therefore, existing research for ISAC has focused on technologies for sharing their infrastructures. For instance, in rural areas where radar sensing is required for monitoring environmental changes, sensing services can be provided using existing mobile communication infrastructure [5]. Also, integrating sensing into wireless local area networks (WLAN) deployed in indoor environments (e.g., smart homes) can enable the provision of ubiquitous services [6], [7]. Recently, ISAC-aided vehicle-to-everything (V2X) communication has been recognized as a promising technique for improving the performance of autonomous vehicles [8], [9].

Traditional ISAC use cases, including the aforementioned research, focus on the cost efficiency that can be achieved by utilizing a common infrastructure for both sensing and communication [3], [10]. However, the mMTC network is envisioned in stark contrast. Verticals strive to enhance security and reliability, even by introducing additional infrastructure. Private 5G has emerged for this purpose, establishing independent mMTC infrastructure using dedicated spectrum [11]. Moreover, unlike radar sensing aimed at immediate detection, there is a paradigm shift towards collecting sensing data over the long term to form big data and adopting a data-driven approach to derive new insights from numerical analysis. Therefore, existing ISAC research is not suitable for mMTC networks in terms of infrastructure properties and sensing mechanisms. Considering the increasing impact of next-generation mobile communication on mMTC scenarios, the traditional ISAC framework must be enhanced to more effectively meet the infrastructure demands (e.g., higher security and reliability) and sensing requirements (e.g., data-driven sensing) of mMTC networks.

In this paper, we propose a novel ISAC framework for mMTC scenarios using private 5G-based IIoT infrastructure. To demonstrate the specific potential of the proposed ISAC paradigm, this paper utilizes the sensor network of a smart factory as an example, which is a representative use case of mMTC and private 5G [12]. The sensor network of a smart factory requires various sensing capabilities (e.g., vibration, current sensor) through IoT gateways [13]. It should be noted that the radar sensing employed in traditional ISAC cannot detect the vibrations and currents of smart factory components. Therefore, ISAC for mMTC should not be limited to radar sensing.

The proposed ISAC framework aligns with existing ISAC research trends, while extending the scope of sensing and shared resource utilization, enabling broader and more efficient applications. While existing ISAC studies focus on signalinglevel integration at the physical layer [3], [14], the proposed framework advances this by integrating sensing capabilities at the packet level within the higher layers of the communication protocol stack. We enable the packet, which was not considered in traditional ISAC frameworks, to convey sensing data, ultimately leading to a novel ISAC framework — dedicated communication for sensing. This represents an enhanced form of communication-assisted sensing, where all communication resources (e.g., time slots, subcarriers) are dedicated to sensing.

To share hardware resources, we can consider a scenario that integrates radar system with private 5G equipment supporting the mmWave band. However, detecting the aforementioned vibrations and currents requires a separate sensing device, which contradicts the device-free sensing pursued in traditional ISAC [2]. Nonetheless, this cannot be avoided in mMTC scenarios, where numerous industrial data must be handled reliably. Therefore, we shed light on the notable necessity of a novel ISAC framework that simultaneously considers devicefree sensing utilizing radar and device-dependent sensing leveraging the 5G communication protocol.

To meet the high reliability and data-driven performance improvements required in the smart factory scenario, the proposed ISAC framework must be able to receive accurate sensing values over an extended period based on stable communication. This capability enables us to derive new insights from long-term accumulated sensing data through an intelligent, data-driven approach (e.g., statistical analysis, AI/ML). To achieve standardized data-driven intelligence for mobile communication, open radio access network (Open RAN, O-RAN) emerged from global collaboration led by the O-RAN Alliance [15]. Thus, the proposed ISAC framework leverages open RAN technology to intelligently manage the sensor networks within private 5G infrastructure, illustrated in Fig. 2.

Fig. 2.

The proposed ISAC framework based on private 5G and open RAN.

It is important that the sensing data packet passing through the IoT gateway traverses the user plane (UP). In the IIoT context, particularly for sensor data, UP data is essential for system monitoring, management, and improving production efficiency. To this end, in this paper, we provide an overview of the O-RAN specifications, specifically the O1 interface, which enables the collection of UP data. Additionally, we propose a novel ISAC framework that facilitates the collection of UP data based on O-RAN architecture.

Fig. 3 shows how the proposed framework can be seamlessly integrated into the traditional ISAC framework. To address the requirements of mMTC scenarios, the private 5G infrastructure serves as the foundation of the hardware layer. The signaling layer manages radar sensing and the communication of short packets generated by sensing devices. Additionally, in the application layer, data-dependent sensing complements radar sensing to leverage the benefits of datadriven solutions. Furthermore, the introduction of the packet layer—a crucial feature of the novel ISAC framework— enhances its overall functionality.

The contributions of this paper are as follows.

· We define a new ISAC framework for the mMTC scenario using private 5G and open RAN.

· We propose a novel O-RAN-compliant ISAC framework for mMTC networks, specifically designed for sensor networks.

· We implement the proposed ISAC framework in real devices, and evaluate its feasibility through experiments on a smart factory testbed with three-phase motors.

The remainder of this paper is organized as follows: Section II investigates private 5G and open RAN. Section III presents the design of the proposed ISAC framework for mMTC scenarios. Section IV evaluates the proposed framework using testbed experiments, and Section V concludes the paper.

Fig. 3.

Integration of the traditional ISAC framework with the proposed framework. We referenced the ISAC layers in [ 3], and highlighted the new features introduced by the proposed framework in red.

II. BACKGROUND

This paper defines a sensor network infrastructure to realize cost-effective and intelligent sensing services for handling sensor data using private 5G and open RAN technologies. This section describes the expected features and advantages achieved by applying private 5G and O-RAN architecture to the proposed ISAC framework for sensor networks.

Compared to the previously established networks for human-oriented communications, the private 5G exhibits distinct characteristics. First, private 5G networks are less affected by mobility because the machines are deployed based on a pre-designed layout. Second, the data packets are typically short (e.g., sensor data) and contain meaningful real-time information. Network administrators can leverage the collected data and develop new services, such as predictive maintenance and real-time monitoring, for reliable management. Third, the operator must minimize the cost of base stations for network deployment, because private 5G cannot generate profit through charging, unlike mobile network operators (MNOs).

The primary objective of private 5G for sensor networks is to collect and manage industrial data efficiently while ensuring cost-effectiveness. However, deploying private 5G with traditional monolithic RAN systems poses challenges due to high costs associated with vendor monopolies. On top of that, traditional RANs lack programmability and standardized data collection and processing capabilities. To address these issues, a novel software-defined RAN architecture that can offer tailored network functions, cost-effectiveness, and support a data-driven approach is necessary.

The key design principles for cost-efficient RAN systems are disaggregation and softwarization. The O-RAN architecture disaggregates the monolithic RAN into radio units (ORUs), distributed units (O-DUs), and central units (O-CU). The RAN functionalities can be optionally divided by the 3GPP NR split specification, and disaggregated RAN elements can implement these functions [16]. The O-RAN architecture leverages split option 7.2x, separating only the functions that are feasible for software-based implementation into O-DU and O-CU. Consequently, it has become possible to establish L2/L3 functionalities through software, and the enhanced computing power supports software-defined RAN to achieve sufficient performance.

Many RAN software solutions have been released as opensource projects in accordance with the ongoing open-source trends in the software industry [17]. Furthermore, the O-RAN architecture defines open fronthaul, enabling the connection of O-RUs and O-DUs through these interfaces [18]. This facilitates the interoperability of equipment from different vendors, allowing different vendors’ O-RUs and O-DUs to communicate through open and standardized fronthaul interfaces. This provides an opportunity for smaller companies to participate in the market, allowing cost reduction on RAN equipment.

Thanks to the open interfaces and open-source software, by purchasing only an inexpensive 7.2x split O-RU, a softwaredefined RAN can be deployed using commercial-off-theshelf (COTS) server. Developing O-DU/O-CU functionalities on a COTS server and then connecting multiple affordable ORUs to that COTS server via an open fronthaul enables the cost-effective deployment of a RAN.

Due to the introduction of intelligence for efficient management, networks can be divided into radio and management side. To provide intelligent management and automation of networks, the O-RAN Alliance defines the service management and orchestration (SMO) framework as the main component of management side [19]. As shown in Fig. 4, the SMO supports data collection and AI/ML-based operations across O-RAN elements through open interfaces. These elements include a disaggregated RAN and components for intelligent data-driven control of networks, which are called RAN intelligent controllers (RIC). The O-RAN architecture includes two types of RICs: non-real time RIC (non-RT RIC) on the management side, and near-real time (near-RT RIC) on the radio side [20]. The non-RT RIC is a logical function internal to the SMO and supports a large-timescale (above 1 second) operation on the radio side. The near-RT RIC can support the control and optimization of disaggregated RAN elements with a smaller latency (from 10 millisecond to 1 second), thanks to its proximity to the RAN.

SMO can control the near-RT RIC and disaggregated RAN with the O1 interface. Non-RT RIC and near-RT RIC are connected via the A1 interface, which enables policy-based guidance for applications in near-RT RIC [21]. The near-RT RIC can collect data from RAN node and provide data-driven control using the E2 interface. Therefore, the introduction of RICs enables the data-driven operation of sensor networks.

It is worth mentioning that originally, RICs aimed to collect only control plane (CP) data to enhance network performance [22]. However, for sensor networks, a much higher requirement for UP data is expected, as mentioned in Section I. Thus, we need a new methodology to collect UP data using O-RAN architecture, which leads to our design of the novel ISAC framework that we propose in Section III.

Fig. 4.

O-RAN architecture with RICs and interfaces.

III. PROPOSED ISAC FRAMEWORK

Having described the O-RAN specification for collecting UP data, we propose the ISAC framework for sensor data applications using RICs and SMO.

In the O-RAN use case analysis report by Working Group 1, application servers are used for retrieving or receiving UP data from the UEs [23]. These servers store UE enrichment information such as radio resource requirements and GPS information. Additionally, for unmanned aerial vehicles (UAVs) applications, they support the capacity to handle substantial volumes of UP data traffic. However, when considering sensor networks, deploying separate application servers alongside RICs to handle short packets is highly inefficient [24].

Working Group 5 has proposed an approach to collect UP data using the SMO and O1 interfaces instead of an application server [25]. The configuration and management of the O-CUCP and O-CU-UP can be accomplished through the O1 interface. The O1 interface supports several management services (MnS) for network operation and management. We introduce four available MnS for sensor networks and propose utilization strategies: provisioning MnS, performance assurance MnS, fault supervision MnS, and file MnS [26].

The provisioning MnS allows the SMO to configure the attributes of the CP and UP of the O-CU. The SMO can handle the security of O-CU-CP and quality of service (QoS) class identifier (QCI), 5G QoS Identifier (5QI), data radio bearer identity (drb-id), slice/service type (SST), and slice differentiator (SD) of the O-CU-UP. These attributes can be utilized to prioritize sensor networks that require high reliability and identify the slicing information of managed nodes.

The performance assurance MnS enables the SMO to receive file-based (bulk) and streaming (real-time) performance data generated from managed nodes. It can be used by a sensor network to report the values detected by the sensors and their inferred performance. The term ‘real-time’ indicates intervals of more than 1 second. Since collecting sensor data at the millisecond level results in excessive overhead, it is more appropriate to gather sensor data at intervals of several seconds. Consequently, the O1 interface can effectively manage both non-time- and time-critical sensor networks.

Service users can perform faults supervision and receive reports on errors and events through the fault supervision MnS. When a sensor detects signs of danger, the fault supervision MnS allows for prompt responses. The file MnS can be applied in conjunction with the performance assurance MnS for integrated usage. The original purpose of file MnS was to facilitate the transfer or download of files from nodes managed by the SMO. In a sensor network, file MnS can transfer large amounts of accumulated data generated by non-time-critical sensors simultaneously. Fig. 5 shows a block diagram of the O-RAN-compliant ISAC framework with these MnS.

Fig. 5.

O-RAN-compliant ISAC framework.

Consequently, in the proposed ISAC framework, UP data representing sensor values are collected only on the management side. As a use case, we can consider utilizing the sensor data collected by the SMO to create AI/ML models for anomaly detection and predictive maintenance in non-RT RIC. Since the near-RT RIC on the radio side is defined by specification to collect only CP data, it cannot gather sensor data directly. Instead, near-RT RIC can implement an application that monitors the outputs of the AI/ML model in non-RT RIC through A1 interface. The application should be capable of delivering response commands in less than 1 second in the event of an emergency on the radio side. It is important to note that the destruction of communication infrastructure due to industrial accidents or natural disasters can lead to significant economic losses [27]. The proposed ORAN- compliant intelligent ISAC framework allows for realtime monitoring of the environment and prompt responses to emerging issues.

Fig. 6 shows the integration process between the proposed sensing functionalities and the traditional ISAC framework— i.e., the configuration of the packet layer. The IoT gateway transmits sensor data to the O-RU of the private 5G network. The packetized sensor data traversing the UP of the O-DU and O-CU is managed by the proposed UP-agent. The UP-agent can capture packets carrying sensor data, process them into a standardized data model, and transmit them to the SMO via the O1 interface. Thanks to the UP-agent, the private 5G equipment can efficiently manage data collected by the IoT gateway. In other words, it can be considered not merely as a receiver of sensor data but as an integrated sensor that manages packetized values. The SMO can access this data and utilize it via a standardized interface (i.e., O1 interface).

Fig. 6.

The integration process for the packet layer of the proposed ISAC framework. It is noted that although UP data is typically transmitted to the internet through the UPF of the core network, the proposed ISAC framework enables the UP-agent to send it to the SMO via a standardized O-RAN interface.

IV. FEASIBILITY EVALUATION

In this section, we introduce our testbed and discuss use cases for the proposed ISAC framework. To implement the testbed in an O-RAN architecture, we develop SMO and RICs that can be deployed in adjacent locations. The RIC is created using open-source software from the O-RAN Software Community (O-RAN SC). In our testbed, we install the SMO and RICs on Intel NUCs with an Intel Core i7-1165G7, 32GB RAM, and Ubuntu 20.04.6 LTS. We use Orange Pi Zero as the IoT Gateway. The private 5G network consists of a baseband unit (BBU) from QCT and a COTS-based core network from LG CNS. The O-RU of our private 5G network operates at a frequency of 4.7 GHz. While this frequency may be atypical for a traditional ISAC framework based on the mmWave band, our selection reflects practical limitations; at the time of establishing our private 5G network, commercially available private 5G devices supporting the mmWave band were not yet accessible in South Korea. However, it is important to note that this choice does not impact our findings regarding the application of data-driven solutions within our proposed ISAC framework. Table I summarizes the parameters for our testbed and simulation.

For empirical validation of our framework, we built a smart factory testbed based on private 5G and three-phase motors. To generate sensor data, we attached vibration sensors to the motors and developed dedicated PCB boards for the collection and storage of this data (Fig. 7(a) and Fig. 7(b)). The actuators can read the sensor data from the PCB board using the Modbus protocol and transmit this UP data packet to the SMO via private 5G and the O1 interface. The sensor data can be monitored remotely using a web server (Fig. 7(c)). The overall architecture of the testbed is shown in Fig. 8.

Fig. 7.

Data sensing system for three-phase motors.

Fig. 8.

Smart factory testbed for the proposed ISAC framework based on private 5G and O-RAN architecture.

To evaluate the feasibility of our framework, we present a predictive maintenance use case for three-phase motors. As shown in Fig. 7(c), we conducted experiments using five old motors, each identified as ID1, ID2, ID3, ID4, and ID5. By leveraging the proposed ISAC framework, we could collect sensor data from these motors via the O1 interface. When some of these motors are nearing faults, the analysis results of the collected sensor data will show clear differences compared to those of normally operating motors. Our data-driven predictive maintenance algorithm can capture these phenomena, predict potential faults, and thus enable reliable operation. We collected vibration data from a motor nearing failure at 5- second intervals over a period of 10 days, and through data preprocessing, we sampled 100,000 data points [28], [29]. To implement data-driven intelligence, we can first determine the probability distribution of the vibration data. Fig. 9 represents a sample from the vibration dataset and its corresponding probability distribution.

To diagnose sudden changes in motor vibration that occur during the early stages of performance degradation, we adopt statistical approaches for analyzing the randomly generated vibration data [30]. In this experiment, we use four statistical measurements: mean, variance, skewness, and kurtosis.¹ The equations for these measurements can be expressed as follows:

¹ Advanced statistical methods for motor fault diagnostics have been extensively studied but are beyond the scope of this paper. Thus, we apply basic statistical techniques to validate the feasibility of the ISAC framework

where [TeX:] $$x, \mu \text { and } \sigma$$ are time series of vibration data, mean of the data and standard deviation of the data, respectively. We calculate the measurements for five old motors and compare the results, which are summarized in Table 1.

Fig. 9.

A sample of our vibration data. Statistical approaches can be applied after determining its distribution.

The variance, skewness, and kurtosis values of motor ID1, which exhibits abnormalities, are significantly higher compared to the measurements of the other motors. This indicates that we can determine the presence of anomalies in motors using data-driven sensing and statistical measurements. Our experiment represents a solid use case for the proposed ISAC framework and can greatly expand the applications of traditional ISAC. In our framework, the SMO can retrieve the vibration data and develop maintenance algorithms. Non- RT and near-RT RICs can use the data distribution to infer the status of the equipment and configure efficient parameter setting.

V. CONCLUSION

In this paper, we propose a novel ISAC framework for mMTC scenarios, specifically termed dedicated communication for sensing, particularly targeting sensor networks. Our framework enables packet-level integration for ISAC, utilizing private 5G and open RAN technologies to address security issues and facilitate data-driven sensor network management. We implement a smart factory testbed comprising a dedicated sensing board, COTS hardware and open-source software. We suggest a methodology for collecting UP sensor data by leveraging an IoT gateway and the O1 interface. In addition, we present a data-driven approach for predictive maintenance and system management. SMO, RICs, and open interfaces form the foundation for the effective operation of a sensor network in the proposed ISAC framework. The proposed framework and testbed can serve as a blueprint for developing use cases for O-RAN and private 5G, enabling an industrial revolution based on 5G communication and data-driven approaches.

While this paper primarily focuses on the system architecture based on O-RAN and private 5G, there are many potential directions for future work. For instance, we can discuss optimal waveform designs and the information-theoretical limits for short packets generated by sensors. Given the use of private 5G spectrum, sensing-assisted physical security could also be an interesting topic to explore. In any future research, we believe that our framework will be a key enabler for reliably and securely handling the surge of communication and sensing data.