Machine Learning Based C-DRX Configuration Optimization for 5G

Abstract:
5G NR defines an energy saving technique called Discontinuous Reception (DRX), which allows a user device to check for incoming downlink traffic intermittently. In this paper, we look at DRX in Connected mode, referred to as C-DRX. Since the number of possible DRX configurations is very large, a common industry practice is to define a fixed set of “good” configurations, which are used depending on the traffic type. In this work, we present a novel approach to optimize the DRX configuration per user according to desired performance goals, or intents, using Machine Learning (ML). Our solution is tested using a 5G NR system-level simulator and the results show that it achieves significant performance gains compared to the baseline configuration according to 3GPP.