Design and development of an autonomous robot based on artificial vision for weed removal in crops in the Junín region of Peru

Beraún-Espiritu M.M.; Moscoso-Paucarchuco K.M.; Llacza-Molina J.D.; Visurraga-Camargo L.A.; Segura-Cuyubamba J.A.; Machaca-Rejas J. · Proceedings of the Annual International Conference on Control, Automation and Robotics, ICCAR (2026) · DOI: 10.1109/iccar69571.2026.11549750

Resumen

In Peru, agricultural intensification and the persistence of pesticide residues above maximum limits in a significant proportion of monitored foods have reinforced the need for technological alternatives that reduce dependence on agrochemicals, including AI-based tools to improve the safety and sustainability of phytosanitary management. In this context, the central problem addressed is weed control, traditionally solved with herbicides and manual weeding, which increases costs, occupational exposure, and environmental risks, especially in production systems in the Junín region. Therefore, an autonomous robot with perception-decision-action architecture was designed and developed, integrating a camera and processing in Raspberry Pi with a YOLOv8 detection model, and a microcontroller for traction control and safe activation of a laser as a selective effector. The results show detection accuracy by class in the range of 87-91%, PID control performance with low overshoot and short response times, and an energy estimate for the system that allows for the anticipation of power requirements and operational autonomy for future field tests.

Instituciones: Universidad Continental; Universidad Nacional José María Arguedas; Universidad Nacional del Centro del Perú; Universidad Nacional Autónoma Altoandina de Tarma