Browse Teach Curate

Path-Loss-Seeded Neural Correction for 4G Radio Metric Prediction

A two-stage supervised-regression architecture in which a machine-learning path-loss model produces an initial RSRP estimate that a correction deep neural network refines using engineered local-environment features. Trained on crowdsourced user-equipment measurements, it predicts the 4G radio-quality metrics RSRP, RSRQ, and RSSI and generalizes across cities through geographically disjoint train/test splits, coming within a few decibels of the path-loss error floor.

B9 Text

Machine Learning for Modeling Wireless Radio Metrics with Crowdsourced Data and Local Environment

This paper presents CRC-ML-4G, a suite of machine learning models for predicting three 4G LTE radio-quality metrics — reference signal received power (RSRP), reference signal received quality (RSRQ),…