By Devaki Chandramouli,  TSG SA WG2 Rapporteur, Nokia

Published December 2025, in Highlights Issue 11

Exciting times are ahead for 3GPP with 6G studies in full swing within all the key working groups. 5G Standalone system and 5G-Advanced features play a pivotal role as a bridge towards the future 6G standard.

Mobile networks provide connectivity to the Internet where apps and cloud services reside. Massive traffic growth is expected due to advancements in AI technology and the future rise of applications such as video, social media, AR/VR apps, multimodal communication and robotics.

Amongst others, recent success of lightweight and trendy smart glasses with AI features is a notable point. While in the past, applications requiring higher throughput for downlink were dominant, in the future, especially with AI the focus will be more on increasing uplink throughput.

When considering how to enforce QoS in cellular networks, especially when there is competition for resources, cellular networks apply similar traffic management principles applied as on the Internet (e.g., based on traffic differentiation, prioritization and shaping). Furthermore, cellular networks offer the ability to support both guaranteed and best effort QoS.

Cellular networks must go one step further, however, as they need to ensure a minimum QoS offering while also managing the variations incurred due to the volatile air interface and the mobility of end users. Going forward, with emerging AI applications, cellular networks should also facilitate the data availability required for AI applications.

5G-Advanced Features

Release 18 (See TS 23.501 - System architecture for the 5G System (5GS); Stage 2) introduced many enablers to enhance the 5G-Advanced QoS framework and support for XR services. These include providing differentiated QoS at the granularity of application data units such as video frames carried in a set of IP packets (PDU sets in 3GPP terminology), and providing additional information to the cellular network on application traffic patterns such as timing and end of data bursts.

The increased application awareness can be used for both QoS and power saving optimizations. These features are building an excellent foundation for optimal treatment of future services in 6G, while at the same time, they help identify those aspects that remain challenging for the current QoS framework.

As the Internet evolves rapidly, e.g., in the area of transport protocols and in policies and business practices, it is crucial to understand the impact these changes will have on cellular networks. As an example, QUIC is an end-to-end encrypted transport protocol providing multiplexing features, meaning traffic with different characteristics or QoS requirements may be carried within the same application traffic flow with no easy way to identify it in the network. Eventually it will be the dominant transport protocol in the Internet, natively supported in cloud environments.

Encryption and multiplexing of traffic without additional identification information will have an impact on MNO offered services:

• Access control – Blocking traffic, fraud prevention, etc.
• Redirection – HTTP, load balancing, etc.
• Parental control.
• Charging policy – Online, offline, zero-rating, etc.
• QoS differentiation – Traffic shaping, bandwidth throttling, etc.
• Content enhancement – Media format resolution, etc.
• Traffic analytics reporting.
• Application and Service awareness.

Release 19 has introduced additional enhancements to improve traffic detection and QoS flow mapping for multiplexed and encrypted traffic.

In the case of multiplexed traffic flows sharing the same IP 5-tuple (destination and source IP addresses and port numbers, transport protocol) with some visible flow identification information in the application protocol such as in (S)RTP, the application server can provide this flow identification information as additional packet filters with associated QoS requirements as part of the “AF session with required QoS” API to the 3GPP network via NEF [1].

The additional packet filters   together with the IP 5-tuple allow the UPF in downlink and UE in uplink direction to uniquely detect each media flow and map it to a specific QoS flow that meets its QoS requirements.

Release 19 also introduced enhancements to enable the application server to explicitly provide per IP-packet (PDU) metadata the UPF can use  to detect PDU Set or Data Burst information and to further provide towards NG-RAN, even for protocols such as QUIC which normally do not carry such information in any network-readable header (that is not encrypted) [1]. Release 19 supports three different methods for obtaining metadata in case of e2e encrypted traffic: Media over QUIC (MoQ), Proxy-UDP-in-HTTP/3 [2] and QUIC-Aware Proxying [3], UDP-option [4].

For MoQ [5], the PDU set information is accommodated via metadata in MoQ. For this, the UPF must support the MoQ relay functionality which allows it to see the MoQ metadata.

Proxy-UDP-in-HTTP/3 [2] and QUIC-Aware Proxying [3] operate by introducing a proxy in AS and by UPF acting as HTTP/3 client establishing connections to the proxy and mapping the desired traffic flows to the proxy connections. In Proxy-UDP-in-HTTP/3 the metadata is included in the HTTP datagrams tunneling the proxied end-to-end UDP payload from AS to UPF. In QUIC-Aware proxying where QUIC packets are forwarded rather than tunneled the metadata is added in the QUIC packet itself by a special packet transform. In both cases the metadata is secured between AS and UPF.

For UDP-Option carrying XRM metadata, UPF also connects to the AS side proxy server as in Proxy-UDP-in-HTTP/3, but the metadata is included in a UDP-Option in UDP datagrams carrying the end-to-end payload tunneling HTTP datagrams rather than in the HTTP datagrams themselves. Security keys for UDP-Option are negotiated using HTTP Connect UDP upgrade token between UPF and AS.

Towards the 6G era

Considering the massive traffic growth, capacity, coverage, reliability, spectral efficiency and high performing radio continue to be critical also in the 6G era.

For easy and smooth migration from 5G to 6G, 6G radio should be able to dynamically share spectrum with 5G. This refers to 5G-6G Multi-RAT Spectrum Sharing (MRSS) as the key tool for migration. The use of MRSS looks promising as 6G radio can be designed  in a way that allows for optimized spectrum sharing spectrum with 5G, in addition to enabling the support of new frequency bands with larger bandwidth using 6G. 5G-6G MRSS is expected to be much more efficient in terms of performance compared to Dynamic Spectrum Sharing (DSS) used for 4G/LTE-5G spectrum sharing.

MRSS performance will not be limited by interference and overheads imposed by LTE common reference signals (CRS), and 5G-6G MRSS can leverage the forward compatible 5G physical layer flexibility. In addition, 6G can also be deployed in newly available spectrum from around 2030 onwards.

It is imperative to be able to support very high data rates throughout the cell to enable excellent user experience, enhanced uplink and support of low latency services.

Furthermore, understanding the boundaries of the current QoS framework is crucial to determine the improvements needed for relevant 6G day-one use cases. With the emerging NextGen-XR and immersive/cloud gaming and multimodal services, one key challenge to enable best-in-class QoS and Quality of Experience (QoE) considering demanding application needs like changing traffic characteristics, low latency and jitter, high data rates, while considering changing wireless network conditions (e.g., high resources needed in the cell edge).

In 5G, radio resource allocation strategies in cellular networks broadly fall into two categories: best effort or strictly guaranteed QoS offering. The first provides no guarantees, which does not always meet the needs of XR or NextGen-XR applications. The second does provide strict guarantees for data traffic (e.g., strict QoS requirements defined and protected by the cellular network for a service like voice or video), but this approach comes with scalability issues for a wide area network by reserving resources in the network even if not used.

For high data rate and interactive services that can adapt to changing traffic characteristics of XR and AI applications, there is a need to study QoS framework enhancements for 6G that allow for a dynamic adaptation of network resources to the need of (existing and new) demanding - yet adaptive - applications.