On a March morning at UC Riverside’s Citrus Research Center, a wheeled robot moved through rows of ‘Washington’ navel orange trees with a demanding task. It needed to collect soil data across an orchard-scale field without manual operation.
For UCR researchers working at the intersection of agricultural robotics and soil science, the mission brought years of related work into the field, showing that a robot could collect detailed soil data across an entire orchard without manual operation.
That progress is detailed in a new paper published in Springer Nature’s Precision Agriculture. The study, “Dense Proximal Sensing of Soil Apparent Electrical Conductivity using Autonomous Field Robotics,” reports fully autonomous, full-scale field mapping of soil apparent electrical conductivity, or ECa, in a micro-irrigated orchard.
The study was conducted by principal investigators Elia Scudiero, associate professor of precision agriculture, associate agronomist and director of UC Riverside’s Center for Agriculture, Food, and the Environment, and Konstantinos Karydis, associate professor in the Department of Electrical and Computer Engineering, in collaboration with UC Riverside graduate students Aritra Samanta, Bhawana Acharya and Dimitrios Chatziparaschis.
Soil ECa is an indirect measure of soil properties such as texture, moisture and salinity. These measurements can help researchers understand how conditions vary across a field and support site-specific crop management. But in orchards, collecting that information can be slow and labor-intensive. Sensors often need to operate near irrigation emitters and under tree canopies, where large equipment and manual surveying can be difficult.
The UCR team addressed that challenge by integrating a wheeled mobile robot, an electromagnetic induction sensor, onboard localization and a planning algorithm designed for orchard-scale operation. The system allows the robot to navigate orchard rows, prioritize important sampling locations and collect dense soil data within a fixed operating window.
“This work pushes forward the field deployment of autonomous mobile agricultural robots via a tight integration of motion planning and control, decision making, and onboard multimodal perception,” Karydis said. “It also offers a crucial step toward scalability in practical use cases, by enabling the autonomous ECa mapping of an entire orchard.”
The paper builds on earlier UCR work in robot-assisted ECa measurement, digital-twin testing and informed sampling. The new study introduces the Best Effort Next-Best-Action Planner, or BE-NBA-P, which extends prior planning work by enabling dense sampling while preserving high-priority sampling locations and accounting for timing constraints.
In field testing, the robot surveyed a site with 248 trees arranged in 31 rows and eight columns. The mission was allotted two hours. The robot completed it in 6,235 seconds, or 1 hour, 43 minutes and 55 seconds.
After the mission, the research team processed the data to create spatial and single-tree ECa measurements. The resulting map shows the robot’s trajectory across the orchard and the ECa values collected across the field, translating a complex autonomous mission into information that can help researchers better understand soil variability.
For agriculture, the work points toward robotic systems that can collect high-resolution soil data with less manual labor and greater consistency. For robotics, it demonstrates how planning, perception, control and decision-making can come together in structured but challenging outdoor environments.
The next steps are to enable the efficient deployment of multiple robots and to consider additional proximal sensing tasks. Those directions could expand the approach beyond single-robot ECa mapping and support broader autonomous data collection in agricultural fields.
The work was supported by National Science Foundation grants CMMI-2046270 and CMMI-2326309; U.S. Department of Agriculture National Institute of Food and Agriculture grants 2020-69012-31914 and 2021-67022-33453; Office of Naval Research grant N00014-18-1-2252; and the University of California Office of the President UC-MRPI M21PR3417.