Artificial intelligence in modern agriculture: recent advances, challenges, and future directions: a systematic review

This is the authors' abstract. We don't add key points for this paper.

Modern agriculture faces mounting challenges from climate change, population growth, water scarcity, soil degradation, and the need to sustain productivity. Artificial intelligence (AI) is transforming the sector by enabling precise, data-driven automation and improved resource efficiency across crop production systems. This systematic review synthesizes 95 peer-reviewed studies published between 2021 and 2025, following the PRISMA framework. It covers AI applications in crop monitoring, smart irrigation, nutrient management, pest and disease detection, yield prediction, phenotyping, robotics, and controlled environment agriculture. Most studies focused on environmental monitoring (23.15%), greenhouse and protected cultivation (21.05%), and pest and disease detection (21.05%). Algorithm analysis indicates a dominance of machine learning and hybrid models (25.26%), followed by Internet of Things (IoT/AIoT) and sensor-based approaches (22.11%). The findings reveal a clear evolution over five years, from isolated, single-purpose models and image-based detection to integrated systems that combine sensing, modeling, and automated control. A notable trend is the growing adoption of explainable and reliable AI, particularly for nutrient prediction, irrigation management, and crop growth optimization. Robotics has also advanced, with applications in fruit harvesting, mushroom grading, monitoring systems, and autonomous agricultural fleets. However, the review identified that most studies remain limited to proof-of-concept or prototype-level evaluations, with insufficient external validation and fewer assessments under diverse real-world agricultural conditions. Recent research now focuses on practical validation, field-level implementation, and integrating AI with IoT, sensing technologies, automation, and decision-support systems. Nevertheless, challenges such as scalability, high implementation costs, limited data availability, poor model generalization across crops and environments, and deployment constraints continue to hinder widespread adoption. Overall, the review highlights strong progress toward integrated AI-driven Agriculture 4.0 systems while emphasizing the need for further real-world validation and scalable solutions for large-scale agricultural deployment.