Background: Traditional weed management in rice farming relies heavily on herbicides and labor-intensive manual practices, which are often costly, inefficient and environmentally unsustainable. The lack of precision in conventional approaches results in excessive chemical use, crop damage and inconsistent weed control. To address these challenges, intelligent robotic systems integrated with artificial intelligence offer a promising pathway toward sustainable and precise weed management. 
 
Methods: This study presents a comprehensive analysis of a modular autonomous robotic platform that integrates advanced mechatronic design with AI-driven visual intelligence for rice-field weed management. Three deep learning models, YOLO v5, v7 and v8, were trained on rice weed image datasets for real-time weed detection and classification. Advanced image-processing algorithms were employed for crop-weed discrimination and row guidance. The robotic platform, powered by an NVIDIA Jetson Nano and planetary gear motors, features a dual-action mechanical subsystem consisting of a rotary weed cutter for physical removal and a cultivator&ndash;incorporator for in-situ biomass mixing. Field experiments were conducted at the Agricultural Farm of Siksha &lsquo;O&rsquo; Anusandhan (SOA) University, Odisha, India. 
 
Result: The robotic platform demonstrated stable real-time navigation, achieving a lateral tracking deviation of approximately &plusmn;2 cm under controlled test conditions and less than 5 cm in field environments, as quantified by measuring the robot&rsquo;s offset from the crop-row centreline at fixed reference intervals. Field trials demonstrated approximately 95% weed control efficiency and less than 2% crop damage. Compared with conventional practices, the robotic system reduced herbicide use by nearly 70% while maintaining stable operation under representative paddy-field conditions. Detection accuracy and field-level weed-removal efficiency are distinct evaluation metrics. The proposed AI-integrated robotic platform demonstrates strong potential for precision weed management in rice cultivation. By combining deep-learning-based vision with a robust mechatronic framework, the system significantly reduces chemical dependency, improves weed-control accuracy and enhances environmental sustainability. This work highlights a viable pathway to scalable, eco-friendly automation in paddy-field agriculture.

Multi-YOLO Comparative Deep Learning-integrated Robotic System for Precision Weed Control in Rice (Oryza sativa L.)