Online Learning of Any-to-Any Path Loss Maps

Abstract

Learning any-to-any (A2A) path loss maps might be a key enabler for many applications that rely on a device-to-device (D2D) communication, such as vehicle-to-vehicle (V2V) communications. Current approaches for learning A2A maps have a number of important limitations, including i) a high complexity that increases rapidly with the number of samples, making the problems quickly intractable, and ii) the inability of coping with a time-varying environment, among others. In this letter, we propose a novel approach that reconstruct A2A path loss maps in an online fashion. To that end, we leverage on the framework of stochastic learning to deal with the sequential arrival of samples, and propose an online algorithm based on the forward-backward splitting method. Preliminary simulation results show a significant decrease in complexity, while its performance is comparable to that of a batch approach.