Machine Learning-Based Modeling of Rainfall Impact on FMCW Radar Performance

This paper presents a machine learning-based modeling framework, designed to quantify the impact of rainfall on Frequency Modulated Continuous Wave (FMCW) radar performance by developing a criterion. the proposed framework encompasses several critical steps, including data collection, data processing, feature selection, performance estimation, and model fitting. Initially, the captured area of radar is divided into multiple units, with each unit sampled on a plane. Six point cloud features are defined to analyze the characteristics of the captured data within each unit. The measured performance is proposed based on the calculated features, which facilitates subsequent machine learning training. The collected data are split into training sets for developing a rainfall criterion and validation sets for testing its performance, with experiment results confirming the effectiveness and precision of the proposed modeling framework.