A heterogeneous sensing system for soil moisture mapping in agricultural environments

Abstract

Soil moisture is a key driver of irrigation management, crop yields, and sustainable food production, yet obtaining accurate, timely, and field-wide measurements remains a challenge for large-scale agriculture. Existing sensing methods suffer from limited spatial coverage, high labor requirements, or prohibitive costs. To address these limitations, we propose and evaluate a heterogeneous sensing system that integrates three complementary platforms: a radio frequency-based wireless sensing network for continuous, large-area monitoring; an autonomous mobile robot for targeted in-situ sampling; and a portable handheld probe for flexible, on-demand data collection. Data from these modalities are fused using Gaussian Process Regression to generate high-resolution soil moisture maps with uncertainty quantification. Field experiments demonstrate that the wireless network reliably tracks moisture trends over days, the robot achieves a 97% autonomous sampling success rate, and the handheld probe delivers sub-1% error after soil-specific calibration. Critically, our integrated system achieves a full-field mapping error of 1.02%, representing an improvement of over 80% compared to the best individual modality. These results demonstrate that our multi-modal fusion approach, which incorporates distributed wireless sensors, robotic in situ sampling, and portable probes, can substantially improve both the accuracy and spatial resolution of soil moisture maps for precision agriculture. • Scalable system for in-situ and wide-area soil moisture monitoring. • Combines RF network, autonomous robot, and handheld probe. • Achieves sub-1% error after soil-specific calibration. • Enables continuous operation across large farms for multiple days.