With the continuous growth in fruit and vegetable yields, the research and application of fully autonomous harvesting robots have become a major focus in the field of agricultural intelligence. However, the coordinated operation across the entire “perception execution planning” workflow of harvesting robots has still faced numerous technical challenges. To address this issue, existing research achievements were systematically reviewed, and the key technologies of fruit and vegetable harvesting robots were comprehensively analyzed, focusing on three core aspects: precise perception of fruit and vegetable targets, adaptive grasping control of harvesting targets, and autonomous harvesting behavior planning. First, starting from the perceptual foundation for autonomous operation, three categories of precise target perception method were reviewed, including deep learning-driven precise perception, multimodal information fusion perception, and active perception with viewpoint planning. These provided essential data support for subsequent execution and planning stages. Building upon this, the end-effector, as the core execution component, was examined in detail. The structural design characteristics of mechanically separated, vacuum suction, and flexible gripping end-effectors were categorized and summarized, and grasping control strategies adapted to different fruit and vegetable types were thoroughly discussed. Furthermore, the development status of robotic manipulators, ranging from industrial robotic arms to agriculture-specific robotic arms and multi-arm cooperative systems, was analyzed, and recent progress in autonomous harvesting behavior planning for robotic manipulators was systematically investigated. Finally, representative construction cases of fruit and vegetable harvesting robots were analyzed. The efficient synergy between perception systems, end-effector modules, and motion planning systems constituted the core imperative for the industrial deployment of fruit and vegetable harvesting robots. The robustness and stability of existing visual perception, grasping control, and motion planning algorithms were found to remain insufficient, while the degree of coordination among these modules demanded further improvement. Future research should concentrate on addressing these technical shortcomings and promoting multidisciplinary collaborative innovation to facilitate the comprehensive upgrading and widespread application of fruit and vegetable harvesting robots.