Efficient Soil Moisture Monitoring without In-Situ Probes: LSTM-Based Bluetooth Signal Strengths Analysis

Soil moisture measurement without damaging soil structure is important in agriculture. Electrical conductivity and microwaves are commonly used for this purpose. Recently, there has been growing interest in using artificial neural networks and time series forecasting to determine soil moisture content. Bluetooth signals, similar to microwave devices, are utilized in these measurements. However, Bluetooth signals have low transmission power compared to dedicated soil moisture measurement devices. This study investigates the possibility of determining soil moisture content using Long Short-Term Memory (LSTM) neural network architecture and Bluetooth signal strengths with 0.001 Watt transmission power, specifically for soil samples with varying pH values. The objective was to assess soil moisture change status directly using Bluetooth signal levels without an in-situ probe. In an experimental study, soil moisture content changing was predicted with a root-mean-square error (RMSE) value of 15% using an artificial learning model based on Bluetooth signal strengths obtained from alkali soil samples. This method eliminates the need for a dedicated sensor, as soil moisture can be reliably measured by tracking signal level changes over time.