Estimating Hidden Information for Self-organization and Self-healing in Modern Wireless Networks

Abstract As modern wireless networks grow in size operators are more interested in self-organizing features that ease management of their infrastructure. Modern networks have a higher number of possible failure points compared to legacy ones thus the operators increasing interest in self-organizing, self-healing mechanisms. This paper proposes incorporating public online social network updates to enhance the ability of an autonomous management system to determine the best course of action in case of cell site failures. This approach has not been considered in the past in any fault detection or self-healing mechanism. The proposal discussed in this article expands the author's previous work in the area and employs distributed probabilistic graphical models (PGMs) to successfully learn and predict which network resources are better suited to recover from a fault. The PGMs take into consideration observable processes like cell handovers or data traffic along with external information coming from online vehicular traffic monitoring systems available through social networks. Then the PGMs are used to determine which operating resources are needed to restore operation over an area for which no up to date information is observable. This new approach was successfully studied and tested in a simulation framework.