Vision-Based Autonomous Navigation: Sensors, Algorithms, Challenges and Future Perspectives

With the rapid development of autonomous transportation and robotics, autonomous navigation has become an important research area. To achieve reliable navigation, advanced sensors and efficient perception algorithms capable of interpreting complex environmental data are required. This paper reviews the development of vision-based autonomous navigation systems, focusing on the sensors, data processing perception systems and current challenges. The characteristics of mainstream sensing approaches, including RGB cameras, infrared cameras and LiDAR, are analyzed through comparisons in their advantages and limitations. Furthermore, major perception algorithms, including visual simultaneous localization and mapping (V-SLAM), point-cloud processing and bird's-eye-view (BEV) perception, are reviewed, putting focus on their processing principles, advantages and computational requirements. Challenges remain in system performance and accuracy, and future development is expected to shift towards improving computational efficiency and enhancing sensing technologies. This paper provides an overview of current progress in vision-based autonomous navigation and insightful directions for achieving more reliable and efficient autonomous navigation systems.