A Polynomial Neural Network Approach for the Outdated CQI Feedback Problem in 5G Networks

Peterson Marcelo Santos Yoshioka, Jose dos Santos, Marcos Falcão, Andson M Balieiro, Elton Alves, Siba Narayan Swain

DOI: 10.14209/sbrt.2024.1571031978

Evento: XLII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2024)

Keywords: Outdated CQI Feedback Problem 5G Networks GMDH Polynomial Neural Network

Abstract

Accurately reporting a Channel Quality Indicator (CQI) value that denotes the current channel condition is fundamental for 5G networks. However, the time elapsed between the channel condition measurement and its effective use by the base station may render the CQI obsolete, negatively affecting the UE communication. This paper proposes a Polynomial Neural Network solution that considers the Signal-to-Interference plus Noise Ratio (SINR) and user context to estimate the updated SINR for translation into a CQI value. It is based on a self-organizing algorithm (the Group Method of Data Handling - GMDH) that combines the concepts of black-box modeling, connectionism, and induction for computer-based mathematical modeling of multi-variable systems and automatically optimizes its structure with minimal analyst intervention. The results show that our solution presents a high level of accuracy and performance similar to the ideal one, with an absolute difference of only 0.001 in both throughput and spectral efficiency metrics, demonstrating its feasibility to address the outdated CQI feedback problem.

Download