With the growing demand for urban cycling, traditional navigation systems have problems such as large visual interference and weak environmental perception. This paper proposes a cycling intelligent navigation system that integrates virtual reality (VR) technology to achieve immersive navigation through multi-sensor data fusion and three-dimensional scene reconstruction. The system adopts an improved A* path planning algorithm combined with a visual-tactile-auditory multimodal interaction mechanism. The experimental part designs three sets of comparative tests: path planning efficiency (MATLAB simulation), user cognitive load (eye tracking), and system response delay (stress test). The results show that compared with traditional navigation methods: In terms of path planning, the improved A* algorithm used in this system reduces redundant path turns by 55.6%, improves path smoothness by 42%, and greatly optimizes path quality. In terms of user attention, users of traditional systems need to look at the device screen for 12% of the time, while this system reduces this proportion to 3% with its multimodal interaction mechanism, significantly reducing visual interference. At the same time, the user’s visual focus is concentrated on the area 20–30 m ahead (accounting for 78%). This system better fits this attention distribution characteristic and reduces the user’s cognitive load. In terms of system response, the average response delay is controlled within 300 ms to ensure that navigation information is conveyed in a timely manner. Studies have shown that VR technology can significantly improve the safety and interactive experience of cycling navigation.