Abstract: With the accelerating development of intelligent coal mining, the importance of vehicle operation safety in underground coal mines has become increasingly prominent. Since most coal mines have been equipped with ultra-wideband (UWB)-based precise positioning systems and wireless communication networks, an active safety early warning system for vehicle based on underground UWB positioning technology and AI-based lane vision technology is developed, considering both economic applicability and technical feasibility. In the application of active safety early warning system in vehicle, the real-time performance of UWB positioning is particularly critical because vehicles travel at relatively high speeds and positioning results need to be directly linked with on-board units. Signal collision is identified as the main cause of lag in UWB positioning output. Existing anti-collision schemes for underground UWB signals mainly rely on technologies such as time division multiple access (TDMA), which avoid signal collisions by randomizing transmission time or allocating time slots through additional channels. However, these methods fail to adapt to high-density tag scenarios, exhibit insufficient flexibility in slot allocation, and cannot dynamically adjust the positioning cycle. To address the limitations of existing approaches, we propose a dynamic anti-collision method for UWB positioning signals in underground coal mines. This method uses UWB positioning substations to centrally allocate dynamic time slots for all tags. Without requiring additional communication signals, it relies solely on UWB communication to dynamically plan the transmission timing and positioning cycle of each tag based on the number of tags present, ensuring no conflict in positioning signal transmission. Field tests conducted in Jinfeng Coal Mine demonstrate that the established vehicle active safety early warning system can identify most collision risks. The proposed dynamic anti-collision method performs satisfactorily in underground static tests and electromagnetic interference tests, and effectively solves the problem of missing tag detection caused by a large number of tags and UWB signal collisions. Nevertheless, it cannot resolve missing detection caused by line-of-sight obstruction at roadway bends (i.e., non-line-of-sight propagation) and vehicle encounters (i.e., dynamic occlusion).