O-Ran Integration Achieves Near-Rt Immersive Volumetric Video with Ultra-Low Latency

Scientists are tackling the significant challenge of delivering truly immersive extended reality (XR) experiences, hindered by the latency issues of current volumetric video streaming methods. Yao Wen, Luping Xiang, and Kun Yang, from the State Key Laboratory of Novel Software Technology at Nanjing University, present a novel framework integrating Open Radio Access Networks (O-RAN) to dynamically manage radio, compute, and content resources in near real-time. Their research formulates a system that optimises rendering quality alongside latency, utilising a sophisticated reinforcement learning agent to achieve a substantial reduction in motion-to-photon latency and improved quality of experience , paving the way for scalable and responsive immersive applications.

The study formulates the rendered-pixel ratio as a continuous control variable, jointly optimising it across the Open Cloud (O-Cloud) compute, gNB transmit power, and bandwidth, all while explicitly balancing resolution, computational load, and latency. This innovative approach leverages a Soft Actor-Critic (SAC) agent, employing structured action decomposition and QoE-aware reward shaping to resolve a complex, high-dimensional control problem.

Experiments conducted on a 5G O-RAN testbed and through system simulations reveal that the SAC agent reduces median MTP latency by over 18%, a substantial improvement over existing methods. The team achieved not only lower latency but also enhanced both mean quality of experience (QoE) and fairness for users, proving the feasibility of RIC-driven joint radio-compute-content control for scalable, latency-aware immersive streaming. This work establishes a new paradigm for delivering high-fidelity XR experiences by intelligently allocating resources based on real-time user needs and network conditions. The research introduces a system capable of dynamically adjusting rendering quality and network parameters to maintain a seamless and comfortable immersive experience, minimising the risk of motion sickness associated with high latency.
The core innovation lies in the joint optimisation of multiple resources, radio bandwidth, transmission power, and cloud compute, within the tight constraints of per-frame rendering, a feat previously unachieved in existing systems. By treating the rendered-pixel ratio as a continuous control variable, the system can finely tune the visual fidelity to match available resources and latency requirements. The use of a Soft Actor-Critic agent, a sophisticated reinforcement learning algorithm, allows the system to learn optimal control policies in a complex and dynamic environment, adapting to varying network conditions and user movements. This adaptive control is crucial for maintaining consistent QoE across a diverse user base and network infrastructure.

Furthermore, the research highlights the potential of O-RAN architecture to support latency-critical applications like XR, demonstrating how its programmable interfaces and distributed control capabilities can be harnessed to achieve unprecedented levels of performance. The integration of the RIC, allowing third-party applications to coordinate radio and compute resources, is a key enabler of this breakthrough. This work opens exciting possibilities for future XR applications, including telemedicine, autonomous driving, education, and entertainment, where low latency and high visual fidelity are paramount. The demonstrated improvements in MTP latency, QoE, and fairness pave the way for more immersive, interactive, and accessible XR experiences for all users.

O-RAN Optimisation Reduces XR Latency

Scientists have achieved a significant breakthrough in immersive volumetric video streaming for extended reality (XR) applications, addressing the critical challenge of ultra-low motion-to-photon (MTP) latency. The system innovatively formulates the rendered-pixel ratio as a continuous control variable, jointly optimising it alongside Open Cloud (O-Cloud) resources, gNB transmit power, and bandwidth under a Weber-Fechner quality of experience (QoE) model. Experiments conducted on a 5G O-RAN testbed and through system simulations demonstrate that the Soft Actor-Critic (SAC) agent reduced median MTP latency by over 18%.

This substantial reduction was achieved through structured action decomposition and QoE-aware reward shaping, effectively resolving a high-dimensional control problem. Measurements confirm that the SAC agent not only lowered latency but also improved both mean QoE and fairness amongst users, showcasing the feasibility of RIC-driven joint radio-compute-content control for scalable, latency-aware immersive streaming. The team meticulously measured latency, QoE, and fairness metrics to quantify the performance gains. The research details an O-RAN-assisted ImViD playback architecture where users consume volumetric content via head-mounted displays (HMDs), with a gNB coordinated by the RIC and supported by the O-Cloud providing integrated communication and computation services.

The system leverages 3D Gaussian splatting as a rendering primitive, recognising its computational demands, and focuses on optimising the real-time playback phase. Data shows that reconstruction, while compute-intensive, is comparatively delay-tolerant, allowing for execution by the service provider without impacting O-RAN compute resources. During. Tests prove that rendering complexity scales with resolution, as the frame time depends on the rendered-pixel count and allocated O-Cloud compute. The RAN then maps the completed frame onto radio resources for over-the-air delivery, ensuring a seamless and responsive XR experience. The framework’s integration of the SMO, Non-RT RIC, and Near-RT RIC enables coordinated control and resource allocation.

O-RAN Control for XR Video Streaming enables dynamic

Scientists have developed an O-RAN-integrated playback framework to improve immersive volumetric video streaming in extended reality (XR) applications. This system addresses the challenge of ultra-low motion-to-photons (MTP) latency by jointly managing radio, compute, and content resources in a near real-time control loop. The core innovation lies in formulating rendered-pixel ratio as a continuous control variable, optimising it alongside Open Cloud resources and gNB transmit power, all while balancing resolution, quality, and latency using a Weber-Fechner quality of experience (QoE) model. Researchers employed a Soft Actor-Critic (SAC) agent, utilising structured action decomposition and QoE-aware reward shaping to tackle the complex, high-dimensional control problem.

Experiments conducted on a 5G O-RAN testbed and through system simulations demonstrated that the SAC agent successfully reduced median MTP latency by a significant margin and enhanced both mean QoE and fairness, validating the feasibility of RIC-driven joint radio-compute-content control for scalable, latency-aware immersive streaming. The authors acknowledge that the current R status to the XR xApp is relatively high due to the overhead of frequent, small VR status updates following image delivery. They also note that rendering time becomes a limiting factor as resolution increases, impacting the user’s immersive experience. Future work will focus on expanding deployments to larger scales, incorporating online learning with safety constraints, and integrating the system with emerging 6G technologies. This research establishes a promising pathway towards delivering high-quality, low-latency XR experiences by intelligently coordinating communication and computational resources, a crucial step for the widespread adoption of immersive technologies. The findings suggest that a unified approach to XR rendering and RAN control can significantly improve the performance and user experience of volumetric video streaming.