Weed control remains a critical challenge in modern crop production, particularly under increasing pressure to reduce chemical inputs and improve environmental sustainability. Recent advances in precision agriculture and robotic systems have enabled site-specific weed management, where interventions are applied selectively based on detected weed locations. While extensive research has focused on improving weed detection algorithms, comparatively less attention has been paid to the characteristics and constraints of different weeding modalities, which ultimately determine field performance. This review presents a systematic analysis of robotic weeding modalities from an actuation-oriented perspective. Specifically, we establish a comprehensive taxonomy of weeding approaches, including mechanical, chemical, thermal, laser-based, electrical, and other emerging methods, and analyze their underlying mechanisms and operational characteristics. Furthermore, we examine the coupling between sensing and actuation, highlighting how different intervention modalities impose distinct requirements on perception outputs. A scenario-based comparison framework is then developed to evaluate the suitability of different modalities across representative agricultural conditions, including pre-emergence control, in-row selective weeding, dense-row crop systems, and large weed situations. Based on this analysis, the limitations of single-modality systems are discussed, and emerging trends toward multi-modality integration and air–ground collaborative weed management are reviewed. Overall, this review shifts the focus from detection-centric approaches to the integration of sensing and actuation in robotic weeding systems and provides a decision-oriented framework to support the design, selection, and deployment of next-generation robotic weed management technologies.
Gao et al. (2026) studied this question.