PGM structures in self-organized healing for small cell networks

As the popularity of dense small cell deployments grows so does the need for self-organizing features. This paper looks at how with hidden, unobservable conditions, probabilistic graphical models (PGMs) can be used to successfully predict which networks resources are better suited to recover from a fault. This results in having a self-healing function that does not require extensive backhaul signaling to operate. The paper first shows how temporal PGMs can be used in the context of fault detection and then extends its proposals to the self-healing realm. The results show how in a majority of cases it is feasible to predict basic characteristics of user distribution and load in a failed site and use this information to determine a path to fault compensation.