Abstract

This study presents a localization system based on a hybrid algorithm that combines Strapdown Inertial Navigation Systems (SINS) and Global Navigation Satellite Systems (GNSS) to estimate the position and movement of cattle in a precision livestock farming (PLF) context. The system is applied to data collected from IoT smart collars worn by dairy cows in a free-grazing system, which includes inertial measurement units (IMUs) and GPS modules. The proposed algorithm is a distance-based approach designed to reduce GNSS usage while maintaining acceptable localization accuracy. A Kalman filter is used to estimate the collar’s orientation from inertial data, enabling the transformation of accelerations from the body frame to the world frame. These corrected accelerations are then integrated to estimate the cow’s displacement. GNSS updates are triggered only when the accumulated distance exceeds a predefined threshold, limiting the drift of the inertial system with minimal energy consumption. The algorithm is compared against a baseline method using fixed-time GNSS updates and is evaluated in terms of localization error (relative to original GNSS position) and energy consumption. The proposed system achieved an error of less than 43 meters for 90% of the cases when using a 50-meter distance-based update and reduced GNSS usage by 96.5%, which suggests a significant decrease in energy consumption. Additionally, it outperformed the fixed-time method by achieving a lower 90th percentile error, particularly when GNSS updates were triggered by distance thresholds greater than 26 meters. © 2001-2012 IEEE.