부산대학교 도서관

In this paper, we apply machine learning (ML) to the prediction of blockage in millimeter-wave (mmWave) communication using camera images and received signal power. Because the received power is attenuated by blockage in mm Wave, radio unit (RU) switching is used to avoid disconnection. In mmWave mobile communication, the RU covers a radius of approximately 50 m, and the user equipment (UE) is assumed to move within that range. Because blockage prediction is based on the radio signal strength indicator (RSSI), it is necessary to use the aspects of the environment that cause changes in RSSI due to the movement of the UE as features, such as the positional relationship between the line-of-sight (LOS) link and the blocking obj ect, the distance between the RU and UE, and the antenna gain based on the angle of arrival. To solve the problem, we propose the use of the top view and environmental parameters as features in ML and a method to generate the top view. In the top view, the obj ect height is represented by a grayscale based on the height difference from the LOS link, and the image area is determined so that the location of the UE and direction of the base station (BS) are unified. The environmental parameters are the distance between the BS and UE and the angle of arrival. Because the proposed top view represents the height and position of the LOS link, the change in RSSI with respect to the UE movement can be predicted when combined with the environmental parameters. The evaluation in field experiments shows that the proposed features can be used to predict blockage in the environment in which the UE is moving.