Current fertilization practices in the country rely on generalized, empirical methods, resulting in inefficient input use and suboptimal crop management. The high cost and limited spatial resolution of traditional soil analysis restrict farmers' ability to optimize fertilization in maize fields. This project proposes an autonomous unmanned ground vehicle (UGV) for characterizing soil fertility and nutrient content in Zea mays L. (hard yellow maize), a crop critical to Peru’s food security.

The system integrates low-cost sensors to measure pH, electrical conductivity (EC), and nitrogen, phosphorus, and potassium (NPK) levels in maize fields. These sensors are calibrated using a machine learning model based on Random Forest, trained with soil samples analyzed at the Instituto Nacional de Innovación Agraria (INIA). To enable sampling, a mechanism was developed that combines a rotary auger—capable of drilling up to 15 cm—and a vertical displacement system that positions the sensor in the soil. This mechanism also ensures operation in compacted soils, protecting the sensor needles. The sampling system is mounted on a mobile platform equipped with GPS-RTK for geolocation and a ZED stereo camera for autonomous navigation and environmental perception.

The UGV collects measurements across the field and transmits data in real time to a digital platform that generates fertilization maps and reports. The goal is to provide farmers with an accessible tool for improved fertilization decision-making in hard yellow maize cultivation. Field tests at Universidad Nacional Agraria La Molina—including comparisons between sensor data and laboratory analyses—demonstrate the feasibility of autonomous operation and the system’s effectiveness in assessing soil fertility under real conditions.

This innovation fosters the adoption of autonomous robotics for sustainable agriculture, with strong potential to increase yields, improve fertilizer efficiency, and support national food security.