Machine Learning Based Threshold Tuning for Optimal Massive MIMO Performance

Abstract

This thesis explores how machine learning can be applied to dynamically set the optimal threshold that determines if a user should be granted Sounding Reference Signals (SRS) resources at cell level. The goal of the thesis is to investigate whether this approach can improve the downlink performance on cell level compared to a static set threshold. Collected network measurements were used to train and evaluate different machine learning models for estimating spectral efficiency across different threshold configurations. Among the two evaluated approaches, XGBoost demonstrated the highest performance. The analysis revealed that features related to long-term network behavior, including the mean User Equipment (UE) SRS Signal-to-Noise Ratio (SINR), the mean path loss of the uplink and the aggregated throughput, contributed the most to the predictions, while short-term variability, the variance of the same parameters, showed less influence. The proposed adaptive threshold selection approach performed better than fixed threshold configurations in 48.3% of the evaluated cases, with a net value of around 0.15 bps/Hz, suggesting that dynamic optimization can improve network performance. At the same time, the thesis identified challenges associated with data quality, overlap between threshold configurations, and disproportion of the data set, which limited the stability and generalization of the model. The findings indicate that machine learning based dynamic threshold optimization is a promising direction for wireless network management and motivates further research on improved data collection and model development.

Popular Abstract

Wireless communication has become an essential part of everyday life.

This thesis investigates whether machine learning can be used to improve downlink performance in Massive MIMO systems.

What exactly does Massive MIMO mean? In Massive MIMO, which stands for Multiple-Input Multiple-Output, the basestation is equipped with multiple antennas. They can be used to serve both single users, called SU-MIMO, and multiple users simultaneously, called MU-MIMO. Massive MIMO enables higher speed and throughput for User Equipments (also called UEs). In Massive MIMO, the base station uses either reciprocity-based beamforming or codebook-based beamforming to direct signals toward specific UEs. Reciprocity is often more desirable than codebook, since UEs can achieve higher speeds due to the extended channel information. Reciprocity-based beamforming relies on Sounding Reference Signals (SRS), which allow the base station to estimate the uplink channel. So, why not use reciprocity at all times? SRS resources are limited and cannot be assigned to every UE at the same time, since this would introduce interference and destroy the good connection. To decide which users should receive SRS resources and therefore also reciprocity, networks today use thresholds of the UEs SRS SINR. SINR measures how strong a wireless signal is compared to interference and background noise.

The baseline for this thesis, is that the threshold is fixed on node level and does not adapt to changing network conditions. As a result, some users who could benefit from SRS may not receive access to it. This thesis explores whether a machine learning model can dynamically adjust this threshold based on predicted spectral efficiency and, if so, whether such an approach can improve overall performance. By collecting, cleaning, and analyzing network data, a machine learning model was developed and evaluated. The results showed improved downlink spectral efficiency in 48.3\% of the evaluated cases compared to using fixed thresholds. These findings indicate that adaptive thresholding using machine learning has the potential to improve wireless network performance and enable more efficient resource allocation. The positive net value of around 0.15 bps/Hz also indicates that the dynamic model successfully outperforms the static threshold. In practice, this has the potential to achieve better network performance for users.

• author: Olofsson, Nora and Lundgren, Ebba
• supervisor: Fredrik Tufvesson
• organization: Department of Electrical and Information Technology
• course: EITM01 20261
• year: 2026
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: Massive MIMO, Machine Learning, Supervised Learning, Random Forest, XGBoost, SRS, Dynamic Threshold, Cell Level, Spectral Efficiency
• report number: LU/LTH-EIT 2026-1152
• language: English
• id: 9229775
• date added to LUP: 2026-06-16 13:22:10
• date last changed: 2026-06-16 13:22:10