Abstract:Visual navigation, as the core technical support for the autonomous operation of mobile robots, its performance directly determines the environmental perception accuracy, positioning and mapping reliability, and the rationality of path planning of mobile robots. This paper systematically reviews the research progress of mobile robot visual navigation, and conducts analysis around the three core aspects: visual sensors, simultaneous localization and mapping (SLAM), and path planning: at the visual sensor level, it focuses on the technical characteristics and applicable scenarios of single-modal, multi-modal fusion visual sensors and new visual sensors; at the SLAM level, it summarizes the technical evolution and performance advantages of traditional geometric SLAM, multi-modal fusion SLAM, and neural implicit SLAM; at the path planning level, it mainly introduces the characteristics and applicable scenarios of traditional algorithms and bio-inspired algorithms. Finally, it summarizes the current challenges faced by the technology and looks forward to future research directions, providing a reference for the further development of visual navigation technology.