AI-Based Digital Twin Refinement and Dynamic Cell DTX/DRX for 6G Networks

다국어 초록 (Multilingual Abstract)

This paper addresses critical challenges in the evolution towards 6G networks by
focusing on two core areas: enhancing data reliability for wireless digital twin environments
and implementing intelligent energy management. Firstly, to overcome the
intrinsic data inconsistencies of ray-tracing simulations and the spatial sparsity of
real measurements, we propose a novel data refinement framework based on the autoencoder.
By training the autoencoder on a small subset of reliable real-world data,
the technique successfully learns the intrinsic channel characteristics and generates
high-fidelity data replicas. This approach dramatically reduced the root mean square
error (RMSE) of the simulated data from 7.1619 dBm to 0.0053 dBm, validating its
efficacy in constructing trustworthy digital twins for high-precision localization services.
Secondly, to improve network energy efficiency, we propose a dynamic control
scheme for cell discontinuous transmission/reception (Cell DTX/DRX) utilizing the
deep Q-Network (DQN) reinforcement learning algorithm. By modeling the problem
as a markov decision process (MDP) based on real cell traffic data, the DQN agent learns an optimal policy that dynamically adjusts Cell DTX/DRX parameters
to balance energy saving and quality of service (QoS) maintenance. Experimental
results confirm that the proposed dynamic policy achieves a superior trade-off, securing
significant energy saving gains over the always-on baseline while maintaining
a high level of QoS, particularly showing strong adaptability across varying traffic
loads. These contributions provide essential technological foundations for ensuring
both the high data accuracy and sustainable operation required by future wireless
networks.

목차 (Table of Contents)

• Abstract
• 1 Introduction 1
• 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
• 1.2 AI-based Network Optimization for Wireless Networks . . . . . . . . 2
• 1.2.1 Digital Twin Data Refinement and Localization . . . . . . . . 2
• 1.2.2 Dynamic Cell DTX/DRX . . . . . . . . . . . . . . . . . . . . 3
• 1.3 Research Contributions . . . . . . . . . . . . . . . . . . . . . . . . . 5
• 2 Autoencoder-based Digital Twin Data Refinement and Localization 7
• 2.1 Overview of Digital Twin for Wireless Networks . . . . . . . . . . . . 7
• 2.2 Construction of 3D Digital Twin Environment . . . . . . . . . . . . 8
• 2.3 Ray-Tracing Simulation and Data Generation . . . . . . . . . . . . 11
• 2.4 Autoencoder-Based Data Refinement . . . . . . . . . . . . . . . . . . 14
• 2.5 Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . 15
• 3 Dynamic Cell DTX/DRX-based Network Energy Saving 18
• 3.1 Overview of cell DTX/DRX in 5G-Advanced . . . . . . . . . . . . . 18
• 3.2 Traffic Prediction Framework . . . . . . . . . . . . . . . . . . . . . . 19
• 3.3 Dynamic Cell DTX/DRX Algorithm . . . . . . . . . . . . . . . . . . 23
• 3.4 Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . 24
• 4 Conclusion and Future Research 27
• 4.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
• 4.2 Overview of cell DTX/DRX in 5G-Advanced . . . . . . . . . . . . . 28
• Bibliography 29
• 국문요지 33
• Acknowledgements 35