Prediction-Based Discontinuous Reception Mechanism for Extended Reality Applications

Extended reality (XR) applications are characterised by their unique multi-modal traffic pattern. However, XR user equipment (UE) presents significant battery-related challenges. Meanwhile, the discontinuous reception (DRX) mechanism was designed and standardised to extend the UE battery life on cellular networks. The DRX mechanism is based on a number of parameters, and studies focused on modelling its behaviour targeted diverse types of traffic on UMTS, LTE, and 5G NR networks. Some of these studies target generating analytical DRX models for specific traffic patterns, while others optimise the DRX mechanism with predictive models for specific applications. With evolving application types on the network, such as XR, there is a need to expand on these previous works to prepare current and future networks for the demands of new-age applications. In this paper, we examine prediction-based DRX optimisation tailored for XR applications. We design a dynamic DRX optimisation algorithm that utilises predictions of traffic inter-arrival times to select dynamic DRX parameters with two performance metrics: energy efficiency and delay. Using publicly available XR data, we utilised state-of-the-art forecasting tools such as Long Short Term Memory and Gaussian Processes models to design our dynamic DRX solution, and we report an improvement over the energy efficiency obtained by the standard DRX mechanism with minimal extra-delay.