Session: ASME Undergraduate Student Design Expo

Paper Number: 168218

Autonomous Weed Control in Rice Fields Using Ugvs and "Segment Anything" Model for Sustainable Agriculture 

The increasing demand for unmanned ground vehicles (UGVs) capable of carrying monitoring devices, environmental sensors, and robotic arms has driven innovation in agricultural automation. In the southern region of Atlántico, Colombia, rice cultivation plays a crucial role in the local economy, ensuring food security and employment. However, weed proliferation significantly reduces crop yields, increasing production costs and necessitating the use of chemical herbicides. Conventional methods often rely on glyphosate, a widely used herbicide in agriculture, which has been linked to carcinogenic effects in humans and severe environmental damage. This compound not only impacts biodiversity in farmlands but also contaminates nearby water sources, posing risks to agricultural communities and ecosystems. Given this challenge, sustainable solutions that minimize agrochemical use without compromising crop quality are needed.

Our approach leverages the Segment Anything Model (SAM) to efficiently differentiate weeds from rice plants, enabling precise and automated intervention. This project presents the design and implementation of an autonomous ground vehicle specifically adapted to the conditions of rice fields in southern Atlántico, Colombia. The system integrates hyperspectral imaging, machine learning, and robotic actuation to identify and eliminate weeds without chemical agents. A machine learning algorithm processes real-time data, classifying vegetation and generating control signals for a robotic arm equipped with a mechanical gripper. This system enables precise and selective weed removal, reducing dependence on chemical herbicides and mitigating their associated risks.

The UGV is designed to navigate between crops through irrigation ditches, ensuring minimal soil disturbance and efficient plant access. By leveraging a vision system positioned approximately 30 centimeters above the ground, the system achieves an optimal vantage point for accurate plant identification. The real-time processing capability of the machine learning algorithm allows the UGV to make instantaneous decisions, ensuring timely intervention and minimizing crop damage.

This research advances UGV-based agricultural robotics by evaluating the real-time weed identification accuracy of the SAM model in Colombian rice fields. In addition to reducing herbicide dependence, this technology helps farmers optimize resources, lower operational costs, and improve crop quality without compromising human or environmental health. The integration of machine learning and hyperspectral imaging ensures high precision in vegetation classification, addressing a major challenge in automated agriculture.

Future developments will focus on optimizing the suspension system to enhance the vehicle’s adaptability to different terrains, refining machine learning algorithms to improve weed detection accuracy, and exploring new applications for this system in other crops. Additionally, research will investigate the scalability of this technology for larger rice fields and its potential integration with other automated farming systems. By eliminating chemical interventions and integrating advanced technologies, this project represents a significant step toward sustainable precision agriculture. It offers an innovative and eco-friendly alternative that increases crop yields, reduces environmental degradation, and enhances economic viability for farmers in the region.

Presenting Author: María Carolina Cantillo Orozco Universida del Norte

Presenting Author Biography: María Carolina Cantillo is a third year mechanical engineering student passionate about innovation, sustainability, and solving real-world challenges. She is particularly interested in thermal sciences, mobility systems, and applied research, working on projects that improve transportation efficiency and reduce environmental impact.

Beyond her academic pursuits, María Carolina is committed to science communication and outreach. She leads initiatives to make engineering more accessible, from creating educational content for social media to participating in student organizations that promote STEM fields. Her work spans from conducting computational simulations in fluid dynamics to developing strategies for sustainable urban mobility.

Driven by curiosity and a problem-solving mindset, she continuously seeks opportunities to apply engineering principles to real-world challenges, striving to contribute to a more efficient and sustainable future.

Authors:

Jorge Isaac Ahumada Riquett Universidad del Norte
María Carolina Cantillo Orozco Universida del Norte
Michelle Andrea Charris Navarro Universidad del Norte
Juana Camila Cruz Moreno Universidad del Norte
Juan David Díaz Fernández Universidad del Norte
Alberto Mario Olivares Ortega Universidad del Norte
Juan Camilo Oñoro Araujo Universidad del Norte
Juliana Castañeda Gnecco Universidad del Norte