Relay Selection in Millimeter Wave D2D Communications Through Obstacle Learning

There has been growing interest in device to device (D2D) millimeterwave (mmwave) communication, due to the promising high speeds and immense amounts of unused bandwidth available. However, mmwaves suffer from unusually high attenuation, through free space, and especially through obstacles. The accepted way to avoid such attenuation is to break up the transmission path into multiple short hops, such that there are no obstacles between nodes. We extend the possibility of using a global positioning system (GPS) based, location aware, centralized approach to the problem of relay selection. Satellite imagery can be used to avoid static obstacles in the transmission path. However, satellite imagery cannot acquire smaller obstacles ones like bushes, trees and signboards. Moreover, depending on the size of the obstacle with respect to the heights of the mmwave base station (BS) and an user equipment (UE), presence of an obstacle does not guarantee obstruction. We propose a simple learning based approach to detect the presence of static as well as dynamic obstacles, without resorting to satellite imagery. We then use this knowledge to efficiently select an appropriate relay for a UE, lowering the chance of allocating an obstacle prone link. Finally we compare our relay selection algorithm with an existing algorithm and show that there is a significant improvement in the quality of link allocation.