Radar and AI-based Soil Moisture Monitoring for Efficient Farm Irrigation

Khan, Ruba; Cheung, Allen

Student Work

Radar and AI-based Soil Moisture Monitoring for Efficient Farm Irrigation

Public Deposited

This study presents a novel approach using Stepped Frequency Continuous Wave (SFCW) radar technology and machine learning models to develop a non-invasive, cost-effective method for soil moisture estimation. Using the Akela AVMU radar system, we collected and processed radar data, which was then combined with ground truth data moisture data collected using the PR2-Probe by Delta- T Devices. Various machine learning models were applied to data the, with Gradient Boosting Regressor achieving the best overall model performance with a test RMSE of 0.408 when predicting soil moisture at the depth of 20 cm and XGBoost Regressor achieving a test RMSE of 0.814 which is the best overall for the depths of 0 to 40 cm combined. Despite challenges like extended model run times, complex data handling, and limited data size, our study achieved significant improvements in non-invasive soil moisture prediction methods. This research helps open avenues for broader applications in agriculture such as ground water level assessment and drought prediction, contributing to sustainable agricultural practices.

This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

Publisher