Research Progress on Visual Navigation and Control Methods for Surgical Robots

Abstract

In the last two decades, surgical robots have moved from specialized platforms to widely used tools in minimally invasive surgery. Their performance increasingly depends on visual navigation, because intraoperative images connect tissue perception, spatial localization, and motion control. Recent work has focused on how cameras, ultrasound, CT/MRI-derived models, optical tracking, and learning-based image analysis can be combined under the time pressure and uncertainty of the operating room. The discussion covers visual perception, multimodal imaging, segmentation, target recognition, three-dimensional reconstruction, image registration, real-time tracking, and vision-guided control in surgical robotics. Particular attention is given to the technical bottlenecks created by occlusion, specular reflection, tissue deformation, limited field of view, calibration drift, and data scarcity. The reviewed evidence shows that visual navigation is most useful when perception outputs are tightly linked to registration accuracy, safety constraints, and closed-loop control rather than treated as isolated image-processing results. Future progress will depend on more robust multimodal fusion, clinically reliable validation, safer intelligent control, and better integration of visual and haptic feedback in dynamic surgical scenes.