Adaptive GNSS–UWB Sensor Fusion for Reliable Localization in Precision Agriculture

Accurate and reliable localization is essential for precision agriculture, where operations such as autonomous navigation, mapping, and agriculture-oriented applications demand centimeter- or even sub-centimeter accuracy. However, satellite-based systems, whether ground-corrected or not, often experience reduced performance in agricultural settings due to canopy cover, multi-path, and Non-Line-of-Sight (NLOS) conditions. This paper presents an adaptive sensor fusion framework that integrates GNSS and Ultra-Wideband (UWB) ranging within an Extended Kalman Filter (EKF). The proposed method explicitly models UWB bias under NLOS, introduces a GNSS health score based on raw measurements and estimates acquired by the receiver for data-driven covariance adaptation, and employs a learning-based approach to tune UWB measurement uncertainty dynamically. Experimental validation in agricultural field settings demonstrates that the adaptive EKF achieves centimeter-level accuracy in open-sky conditions and maintains 2D horizontal RMSE below 6 cm in the partially obstructed (NLOS) field tests, outperforming standard fusion approaches by more than 40% in RMSE. The results demonstrate the potential of adaptive multi-sensor fusion to deliver robust and cost-effective localization for agricultural automation.