The complicated ocean environment results in the random time-space-frequency variation of underwater acoustic channels，especially for the shallow water channel that is seriously affected by interface and human interference，which causes great difficulties to underwater acoustic communication. In recent years，the rapid development of machine learning technology in the field of artificial intelligence provides a new idea for improving the reliability of underwater acoustic communication in complex environment. However，due to the complexity and variability of underwater acoustic channels and the lack of universal model，underwater acoustic communication data samples are seriously insufficient from the perspective of machine learning，and the data collected by traditional single and short-term underwater acoustic communication experiments cannot fully represent the spatial and temporal characteristics of underwater acoustic channel. Through the long-term deployment of underwater acoustic communication and signal acquisition equipment，it is expected to enrich communication data to a certain extent and provide data support for the research of artificial intelligence in underwater acoustic communication under typical sea areas. In this paper，a shallow-water acoustic communication buoy is designed and implemented. Equipped with an underwater acoustic communication system，it can realize data collection and performance evaluation of long-time shallow-water acoustic communication. At the same time，it can provide remote status display and function control functions based on the Beidou System. The shallow sea experiment in Xiamen Port shows that the underwater acoustic communication data collected by long-time buoy is of positive significance to the evaluation of underwater acoustic communication performance in shallow sea channel and the correlation analysis of channel characteristics and communication performance，which lays a good foundation for further research.
