Exploration of XR traffic with IPACT and IPACT-GE DBAs for immersive communication

Abstract

Visual engagement is rapidly shifting from two-dimensional (2D) to three-dimensional (3D), driving the growth of immersive eXtended Reality (XR) services. The increasing availability of XR devices accelerates this transformation. As wireless networks struggle to meet XR's stringent bandwidth and delay requirements, optical access networks, particularly passive optical networks (PONs), offer a strong alternative. This paper evaluates the “interleaved polling with adaptive cycle time (IPACT)” and “IPACT-grant estimation (IPACT-GE)” dynamic bandwidth allocation (DBA) schemes in a 10-gigabit Ethernet PON (10G-EPON) under realistic XR traffic defined by optimal frame rate (rf) and data rate (rd). The analysis examines how the maximum allocated window size (Wmax) affects jitter (J) and end-to-end mean packet delay (de2e) and shows how the estimation factor (α) in IPACT-GE and credit factors (X and K) in IPACT control system responsiveness. Both schemes effectively manage upstream bandwidth across varying XR loads. Building on these insights, this paper proposes an “exponential variance-predictive (EVP)” DBA scheme that proactively allocates bandwidth. EVP-DBA combines variance-assisted adjustment with an exponential moving average (EMA)-based traffic prediction to forecast buffer expansion, enabling the optical line terminal (OLT) to allocate adaptive headroom based on both mean traffic and arrival uncertainty. By integrating XR frame segmentation with an online, targeted scheduling cycle, EVP-DBA effectively reduces pre-transmission delay (dpre), optimizes cycle time allocation, and stabilizes end-to-end delay (de2e). The analysis identifies a trade-off between grant wastage (Gwaste) and de2e under varying loads. Simulation results demonstrate that EVP-DBA achieves lower delay, higher grant utilization, and improved bandwidth efficiency, making it well-suited for high-rate, delay-sensitive XR applications in the next-generation optical access networks. © 2026