5G-Advanced Architecture Evolution: An Analysis based on Rel-18 Progress

Since 5G has been commercially launched for more than three and a half years, its deployment and usage have continued to increase.

By the end of November 2022, there were 2.287 million 5G base stations in China. The China Academy of Information and Communications Technology (CAICT) predicts that by 2025, the 5G penetration rate in China will exceed 60% for individual users and 50% for large industrial companies. For example, by the end of 2022, China Mobile had already developed 16,000 industrial 5G use cases, more than 4,400 premium private 5G networks, and over 1,200 5G smart factories. China Mobile has deployed a private 5G network for CATL spanning six provinces, with a total coverage of more than five million square meters, the largest of NPN (Non Public Network) in the industry.

I. Rel-18 progress

Generally, 3GPP publishes one release of 5G technical specifications every two years, starting from Rel-15. The timeline of each release is revised following several rounds of discussions due to the inclusion and the content of new features. Recently the timeline for 5G releases was also impacted by COVID-19. Rel-17, for example, is the first release that was completed remotely via online sessions due to travel restrictions. At the 3GPP SA plenary meeting held in December 2022, 3GPP fine-tuned the Rel-18 timeline. As the System Architecture and Service Working Group 2 (SA2) needed more time to complete some feature definitions, 3GPP extended the Rel-18 freeze date by three months. The final stage-2 specification will be finalized by June 2023; protocol design will be completed by March 2024; and Open API and ASN.1 will be released in June 2024, which will serve as a development guidance for vendors.

Figure 1 Timeline of 3GPP 5G releases

Rel-18 is the first release of the 5G-Advanced standard, which acts as a bridge between 5G and 6G, and Rel-18 and Rel-19 technologies will facilitate the incorporation of potential 6G technologies into 5G standards. Certain Rel-18 features, for example, are designed to support emerging requirements of existing network deployment, while others reflect the evolution from 5G to 6G.

II. Rel-18 architecture and typical technologies

1. Vertical markets: Further improvement of local networking capabilities

Since the commercial deployment of 5G networks, edge computing has emerged as 5G's most successful network capability targeting vertical industries. According to a survey on 5G network industrial usage, more than 80% of enterprise requirements are related to edge computing. This is attributed to its simple deployment requirement and the obvious gains (local service access) it delivers. The related 5G LAN, which is defined in Rel-16, also became one of the most promising features introduced to production networks. Rel-18 has addressed some challenges that edge computing now faces in deployment and scalability by supporting local access for roaming UE (user equipment), inter PLMN edge server access, and the selection of common edge service server.

There are two major types of non-public networks (NPNs). The first is Standalone Non-Public Networks (SNPNs), which are private networks that offer powerful functions. It enhances idle and connected mode mobility and connectivity via non-3GPP networks (e.g., WLANs). The second type is Public network integrated NPNs (PNI-NPNs), which are non-standalone private networks built upon public networks that provide enhanced capabilities through the enhancement of network slicing.

Rel-18 further enhances network slicing. It enables third-party usage of slices based on the maximum number of UEs accessing the network, load prediction, and time restriction. Unlike previous releases, network slicing in Rel-18 has drawn relatively less attention from enterprises. A major reason is that network slices are deployed a bit slower than expected and the current solutions for slice management and roaming support are too complex. Further studies are needed to determine how to simplify network slice deployment.

2. Individual users: Systematic support for immersive XR and multimedia services

To better support immersive XR and multimedia services, Rel-18 working groups have made systematic efforts in RAN, network architecture, and codecs. For end-to-end systems view, multiple working groups including RAN and SA have performed enhancements for XR and multimedia services.

The RAN working group is conducting research on XR, with a focus on the following three areas:

(1) XR service awareness, based on awareness of XR service features in the uplink and downlink characteristics and application-layer parameters to assist RAN scheduling

(2) Power savings for XR services, based on data characteristics such as periodicity and reliability requirements

(3) Capacity improvements, covering improvements to resource allocation and scheduling for XR services to support multi-modality flows and reduce jitter

Figure 2 QoS mechanism with PDU Set Concept

System architecture enhancement includes the coordinated transmission of multi-modality flows for XR and tactile services, how network exposure supports interaction between 5GS and applications, QoS and policy enhancement, and UE power saving. Special attention will need to be given to the coordination of multi-modality flows and the new QoS mechanism.

In terms of coordinating multi-modality flows, problems to be addressed by Rel-18 include:

• How to enable interaction between applications and 5GC.
• Application-layer synchronization and QoS policy coordination for the multi-modality flows (video, audio, and tactile services) of a single UE.
• Whether QoS policy coordination between multiple UEs is required.
• QoS policy coordination and adjustment enhancement of multi-modality flows of multiple UEs and coordination between applications.

Other tasks include marking data packets of different levels of importance within the same XR service and implementing a consistent network policy assurance mechanism for multi-modality flows (video, audio, and tactile data).

Current networks provide QoS guarantees with QoS flow as the finest granularity, and transmit data at the granularity of Packet Data Unit (PDU). This means that they cannot provide a QoS guarantee for video services at the granularity of frame. This is an issue that needs to be addressed for XR and multimedia services. Specifically, the parsing of B-frames and P-frames depends on the successful transmission of I-frames. Thus, it is important to ensure the integrity of frame-based information transmission. To address these issues, 3GPP introduced the PDU Set concept, which provides QoS guarantees at a finer granularity and ensures frame integrity during packet transmission.

Based on the above work, GSMA is expected to introduce separate slicing requirements for XR and multimedia services, and the standard types of XR slices will be specified in 3GPP standards accordingly.

3. Usage of AI in networks is being explored

Two Rel-18 work items relate to the usage of AI in networks.

The first is the enhancement of the Network Data Analytics Function (NWDAF), which focuses on:

(1) Combining federated learning and mobile communications technologies to develop a commercial solution under the data privacy protection consideration. This will maximize the value of the massive amounts of data within telecom networks.

(2) Using execution results as an input in the model training and inference phase to improve the training model and make analyses more accurate.

(3) Coordination with the Management Data Analytics Function (MDAF), Using the intelligent analysis result from the network management domain and the additional analytics of the network input to improve accuracy.

The second work item is 5G network support for AI/machine learning (ML)-based services, which focuses on:

(1) 5G-assisted AI/ML model distribution, transmission, and training to serve diversified applications such as video and voice recognition and robot control.

(2) Network information exposed to AI/ML applications and associated QoS and policy enhancements.

Figure 3 AI model distribution on device, edge, and cloud

The functionality of NWDAF was defined in Rel-15, but it has not yet been commercially deployed on networks. Further study is required on how to better use network data and combine AI with 5G networks to create tangible value for operators or users.

4. Further enhancement of space-air-ground integrated communications

The research and standardization of new 5G-based satellite communications systems have also been drawing wide attention in the industry in recent years. Satellite communication standards have been discussed by 3GPP, ITU-T, and IETF. Satellite communication systems are part of non-terrestrial networks (NTNs), an umbrella term defined by 3GPP that refers to all non-terrestrial flying objects, including high-altitude platform systems (HAPS), air-to-ground systems, and unmanned aerial vehicle (UAV). The characteristics of NTN systems include long distance, fast mobility, and wide coverage.

Figure 4 User plane functions (UPFs) deployed on Geostationary Orbit (GEO) satellites

In terms of networking architecture, Rel-18 supports satellite-based edge computing via UPFs on-board and enables two local switch models:

(1) A single session management function (SMF) with the UPF on-board functioning serving as the Uplink Classifier (UL-CL)/Branching Point (BP) or as the locally deployed PDU Session Anchor (PSA) UPF (the SMF instructs the UPF on establishing the N9 tunnel).

(2) A UPF on-board functioning as the PSA UPF (rules for the N4 session in the UPF are instructed by the SMF).

As 3GPP cellular network protocols and standards become widely adopted, we can quickly expand subscribers of non-territorial networks by taking advantage of the 5G ecosystem and existing subscribers of mobile networks. This can dilute the effort of construction, maintenance, and promotional costs of space-air-ground integrated networks. The integration of networks involves the integrated value chain, which requires both technical and industry support. Technologies are further converging from transparent packet forwarding to deploy base stations and UPFs on satellites. Some architectural changes have been preliminarily discussed, but they are limited to GEO satellite deployment scenarios. Mobility for low-earth orbit (LEO) satellites and network functions convergence will bring topology changes, which can be discussed in 6G standards.

3GPP Rel-19 will define new scenarios for satellite communications, including support for IoT devices for storing and forward operations in discontinuous links, positioning enhancement for satellite access independent with Global Navigation Satellite System (GNSS), and communication between UEs within the same satellite. For users, the immediate change will be direct access to satellite networks via their phones. Preliminary prototypes have been verified to achieve this goal. For commercial terminals, many cell phone manufacturers began releasing phones that support satellite communications since 2022. These products have attracted wide attention across the industry, although they are generally only used for emergency scenarios.

III. Four Rel-19 highlights

Rel-19, the second release of 5G-Advanced, will continue to explore new network service capabilities. 3GPP SA1 has already initiated 13 projects to study related requirements and scenarios, and is expected to complete the definition of requirement standards by November 2023. These projects will study new items such as integrated sensing and communication and the metaverse applied in mobile networks.

1. 5G-enabled XR and media services will be further enhanced to support the metaverse. The study on XR and media services in Rel-18 was derived from the "tactile and multi-modality communication services" defined by 3GPP SA1. With the rise of the metaverse, some of the requirements that seemed ahead of market requirements have become more acceptable to standard organizations.

The Localized Mobile Metaverse Services study item under SA1 studies how to:

• Offer the shared and interactive user experience of local content and services to users in the proximity or remotely.
• Support interactive XR shared among multiple users in a single location.
• Identify users and other digital representations of entities interacting within the metaverse service.
• Acquire, use, and expose local physical and digital information to enable metaverse services.

These are part of a next-generation communication infrastructure that will underpin the metaverse, which is a hybrid of the physical and digital worlds.

Figure 5 Mobile metaverse supporting immersive gaming and live shows

2. Integrated sensing and communication (ISAC) is a 6G technology that will be incorporated into 5G standards. China's IMT-2030 (6G) Promotion Group and IMT-2020 (5G) Promotion Group have both launched research projects on ISAC. Chinese companies are important contributors to this technology. As a new network capability, its usage scenarios, key technologies, privacy and security, and other aspects require significant work before it can be considered mature enough to be deployed. Improvements in radio access technologies are crucial to make ISAC reality, and there are still a number of other core technical problems that need to be overcome. For example, high-frequency bands enable higher sensing accuracy but provide poor coverage. While, low-frequency bands cover a wide area but provide low sensing accuracy. Despite these issues, ISAC remains attractive to operators and vendors, and will become a trending tech in future.

3. A systematic approach is needed for low-carbon designs. Energy saving is another key issue that the industry focuses on. Unlike the traditional design approach that considers energy saving individually, like in the domains of terminals or wireless networks, new systematic approaches will consider energy saving across the entire communications system. In addition, information about the source of energy (e.g., solar energy or electricity from fossil fuel) in networks can guide service usage. For example, when the cost of energy is low, networks can run energy-intensive services. Passive IoT and ambient IoT are good examples of how to minimize power consumption of UEs by harvesting power from the surrounding environment (e.g., radio waves, solar, wind, vibration, and heat), which allows for battery-less or limited storage devices. When a fire occurs, passive IoT devices can be used to track personnel without power source concerns. As the world aims to reach a carbon emissions peak by 2030, the energy consumption of 5G-Advanced and 6G systems will be at the core of sustainable development.

4. Mobile computing networks can be built by integrating computing and 5G-Advanced. Can computing power become a service like network connectivity and emerge as the second source of revenue growth for operators? The answer to this question will largely determine whether operators can evolve from communication service providers to information service providers. Computing force networks have become a buzzword in China's communications industry over the past two years. They have attracted special attention from many Chinese operators, including China Mobile, which has made computing networks part of its company strategy. To make mobile computing networks reality, we need to seek innovation in network architectures and make further advances in computing technologies, including computing power measurements and computing task decomposition.

IV. Summary and outlook

Since the advent of 5G, the mobile communications industry has adjusted its focus from communications services to information services. This trend has been driven by the need to build a thriving mobile communications industry and digital society. As the first release of 5G-Advanced, Rel-18 will support both live network deployment requirements and future network evolution. Many Rel-18 projects are currently underway, including 28 projects under 3GPP SA2. This paper has touched on just a few of them.

There are also some questions relating to standardization that we need to answer: How can we effectively manage the number and quality of projects in a limited time with a limited workforce? This problem has drawn the attention of experts from various standards organizations, and related stakeholders have discussed the timeline and operation modes for Rel-19 projects.

It is more and more encouraging that this problem can be solved in Rel-19.