Agentic AI for Smart and Sustainable Precision Agriculture

Parvathaneni Naga Srinivasu¹Aruna Pavate²G. JayaLakshmi³Jana Shafi⁴Jaeyoung Choi^5*Muhammad Fazal Ijaz^6*

• ¹Amrita Vishwa Vidyapeetham School of Computing Coimbatore, Coimbatore, India
• ²University of Mumbai, Mumbai, India
• ³Department of Information Technology, Siddhartha Academy of Higher Education, Vijayawada, 520007, India, Vijayawada, India
• ⁴Prince Sattam bin Abdulaziz University College of Computer Science and Engineering, Al Kharj, Saudi Arabia
• ⁵Gachon University, Seongnam-si, Republic of Korea
• ⁶Torrens University, Melbourne, Australia

The current study explores how Agentic Artificial Intelligence (AAI) can ensure smarter and more sustainable farming. The study presents a framework that outlines the integration of AAI in the precision agriculture (PA) domain, which could assist the farmers in making much better decisions, using fewer resources, and reducing their environmental impact. The paper outlines a practical plan on how this system could be set up on farms, including how different devices could communicate with each other for better informed decisions. With consistent monitoring of the agricultural farms, the system would assist in increasing the crop yield by taking timely actions concerning the fertilizers, pesticides, and nutrients. The current study includes a case study on tomato disease classification, demonstrating the feasibility of the approach using DenseNet121, MobileNetV2, EfficientDet-D0, and YOLOv8 models as local models in the federated learning environment. The global model achieved an accuracy of 96.4%, outperforming individual client models, while DenseNet121 and MobileNetV2 reached accuracies of 95.0% and 93.9%, respectively. For object detection of weed species, EfficientDet-D0 achieved superior performance with 𝑚𝐴𝑃@0.5 of 0.978, average Precision of 0.865, and observed F1-score of 0.961, compared to YOLOv8 with 𝑚𝐴𝑃@0.5 of 0.956 and 𝐹1 −𝑠𝑐𝑜𝑟𝑒of 0.935. The study also presents the Strengths, Weaknesses, Opportunities, and Threats(SWOT) analysis, that highlights the strengths and deployment constraints of AAI. This fundamental analysis of AAI integration over FL would lay the roadmap for future research, outlining opportunities for developing more resilient intelligent precision agriculture systems.