From Reinforcement Learning to Agentic AI in Crop Management: A Scoping Review with Prospects for Smallholder Maize Systems in West Africa

Project Description for OSF: This project presents the pre-registered protocol for a scoping review titled "From Reinforcement Learning to Agentic AI in Crop Management: A Scoping Review with Prospects for Smallholder Maize Systems in West Africa," to be submitted to Computers and Electronics in Agriculture. Purpose The review systematically maps the current state of reinforcement learning (RL) and agentic artificial intelligence (AI) applications in crop management, with the dual objective of identifying methodological trajectories and critical gaps in the existing literature, and proposing a conceptual framework for an agentic AI system tailored to smallholder maize management in West Africa. Despite rapid advances in RL-based crop management, particularly in nitrogen use efficiency optimization, existing systems predominantly address single management variables under temperate, high-input assumptions, with limited applicability to tropical smallholder contexts. The emergence of agentic AI, in which large language models and multi-agent architectures act autonomously across multiple data sources and decision domains, presents a transformative opportunity to integrate nutrient management, pest control, and farmer-facing advisory functions into a unified intelligent system. Scope The review covers peer-reviewed literature published between January 2018 and December 2025, searched across Scopus, Web of Science, IEEE Xplore, and Google Scholar. It is conducted in accordance with the PRISMA Extension for Scoping Reviews (PRISMA-ScR) framework (Tricco et al., 2018) and the five-stage methodology of Arksey and O'Malley (2005). Expected Outcomes The review is expected to produce four key outputs: a peer-reviewed scoping review manuscript; a PRISMA-ScR compliant flow diagram of study selection; a data charting summary table of all included studies; and a conceptual framework for agentic AI-driven integrated crop management, combining multi-nutrient dynamics (N, P, K), Fall Armyworm integrated pest management (IPM), and smallholder-appropriate interfaces, applicable to West African maize systems. The findings are intended to directly inform future experimental research and graduate-level investigation at the intersection of AI and tropical smallholder agriculture.