Robotics and Computer Vision in Precision Agriculture: A Systematic Review of Applications and Technology Readiness

This paper presents a systematic review of recent advances in robotics and computer vision for precision agriculture, analyzing 118 peer-reviewed articles published between 2021 and 2025. The objective of this study is to synthesize technological advances, identify research gaps, and provide a roadmap for the maturity of agricultural robotics. The research contribution involves the development of a structured taxonomy that links agricultural tasks, robotic platforms, sensing modalities, and AI models, the proposal of a conceptual pipeline from perception to actuation, and a critical synthesis of technology readiness levels (TRLs) and methodological biases. Searches were conducted in specialized engineering and robotics databases, following PRISMA guidelines and applying explicit inclusion and exclusion criteria. From 440 initial records, 118 articles were selected after screening. Extracted variables included platform configurations (aerial, ground, and fixed systems), end-effector designs, sensor types, learning architectures, validation environments, and reported performance metrics. Results showed that harvesting and manipulation dominated the literature, followed by weeding, phenotyping, and counting. RGB-D cameras and YOLO-based detection models were the most prevalent, whereas LiDAR and multispectral sensors were increasingly used for navigation and diagnostics. Most systems remained at intermediate TRL stages (3-6), reflecting limited readiness for real-world deployment. This review concludes that advancing agricultural robotics requires standardized evaluation protocols, multi-season public datasets, and collaborative efforts to accelerate prototypes into large-scale implementation.