抄録
The rapid transition toward Agriculture 4.0 has exposed a fundamental question for autonomous agricultural machinery: how can heterogeneous environmental information be transformed into reliable, adaptive, and scalable machine autonomy under complex and dynamic field conditions? This review addresses this question by systematically synthesizing the technological foundations, application domains, performance advantages, and adoption constraints of autonomous agricultural machinery and by developing an integrated framework organized around observation, heterogeneity, and infrastructure. Rather than classifying technologies solely by machine type, the review examines the coupling among multimodal perception, AI-enabled decision-making, autonomous navigation and control, precision execution, and multi-machine coordination, with particular attention to the challenges of environmental uncertainty, real-time decision-making, interoperability, and system-level scalability. The analysis further identifies a progressive transition from isolated task automation toward data-driven, adaptive, and networked agricultural autonomy, while emphasizing that improvements in individual components do not necessarily translate into system-level performance without compatible computational, communication, mechanical, and institutional infrastructure. On this basis, future research priorities are synthesized around robust multimodal perception, adaptive intelligence, interoperable multi-machine coordination, lifecycle sustainability, and accessible deployment models. The proposed framework reframes autonomous agricultural machinery as an interconnected physical–digital infrastructure in which observation characterizes agricultural states, heterogeneity governs adaptive decision-making, and infrastructure enables reliable execution and coordination, providing a conceptual basis for understanding and advancing scalable agricultural autonomy.
