Reinforcement Learning Based Rate Splitting for Minimizing Age of Information in Finite Blocklength Systems

Date

2025-8-27

Author

Kaya, Enes

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

141
views

0
downloads

Cite This

Age of Information (AoI) is a fundamental semantic performance metric used to evaluate data timeliness in systems that require instantaneous transmission of small state update packets. This study investigates the problem of minimizing age of information in multi-user wireless networks operating in the finite block length (FBL) regime, a fundamental requirement for low-latency applications. While Rate Splitting Multiple Access (RSMA) provides a powerful and flexible framework for managing interference, the joint optimization of its parameters, such as precoding, power allocation, and scheduling, to ensure information freshness is analytically unmanageable in complexity. To overcome this complexity, a Deep Reinforcement Learning (DRL) framework is proposed. We formulate the AoI minimization problem as a Markov Decision Process (MDP) and use the Twin Delay Deep Deterministic (TD3) policy gradient algorithm to learn near-optimal policies for resource allocation. The key contributions of this work include the development of DRL-based solutions that jointly optimize power allocation and precoding vectors in Multi-User Multiple Input Single Output (MU-MISO) broadcast channels. We first develop a framework for Non-Orthogonal Multiple Access (NOMA) and then extend it to the more general and complex RSMA scheme, where we design a DRL architecture to manage the increasing parameter space. Our research covers both perfect and imperfect CSIT scenarios, demonstrating the robustness and adaptability of the proposed approach. Simulation results confirm that our DRL-based method significantly improves information freshness over benchmarked schemes.

Subject Keywords

Age of Information, Reinforcement Learning, Rate Splitting Multiple Access, Finite Blocklength

URI

https://hdl.handle.net/11511/116221

Collections

Graduate School of Natural and Applied Sciences, Thesis