Enhanced drift self-calibration of low-cost sensor networks based on cluster and advanced statistical tools

Abstract

The escalating use of low-cost sensors in environmental monitoring demands effective management of sensor drift and errors. To tackle this challenge, lightweight calibration techniques are essential for accurate sensor readings. This paper presents a novel calibration method for low-cost sensors, prioritizing the detection and correction of sensor drifts. The approach incorporates clustering for efficient task distribution and leverages Inverse Distance Weighting (IDW) for calibration precision. Also, advanced statistical tools, including the Two-Sample Kolmogorov–Smirnov test (TSKS test) and Exponential Moving Average (EMA), assess and enhance sensor data stability based on historical measurements. Additionally, the Root Update Estimator (RUE) is utilized to adjust predicted values and improve the model’s adaptability to changes in the underlying data distribution. Extensive simulation experiments using an Intel Berkeley Research Laboratory (IBRL) dataset affirm the method’s effectiveness in swiftly addressing sensor drifts. These findings advance low-cost sensor calibration, boosting reliability and precision in environmental monitoring.