I. Introduction
Recently, extensive research and standardization efforts have been channeled toward the definition of the paradigm of self-organizing networks (SONs), which aim at achieving a substantial reduction in capital and operational expenditures (CAPEX and OPEX) by reducing human involvement in network operational tasks, while optimizing network coverage, capacity, and quality of service [1], [2]. SONs aim to replace the manual operational processes that have been executed in legacy cellular networks since their conception, such as configuration, postdeployment optimization, and troubleshooting, with autonomous functions called SON functions, such as self-configuration, self-optimization, and self-healing [1]– [3]. The main task within the self-healing functionality is autonomous cell outage detection (COD) and its compensation. Traditionally, cell outages have been manually detected. In some cases, cell outage can be detected by the manual analysis of fault alarms at the operations and maintenance center (OMC), whereas other detections require site visits or drive testing. This is an expensive process. In addition, it may take hours or days for the cell outage to be detected, thus resulting in pronounced reduction in capacity and quality of service and coverage gap [4], [5]. Once detected, the outage is compensated in an ad hoc and manual fashion, making the whole process extremely inefficient and unreliable. With an increasing scale of networks, automatic detection and compensation of cell outage has become a necessity, and it has been included in recent Third-Generation Partnership Project (3GPP) releases [6]. Therefore, the SON paradigm aims to replace these manual tasks with an autonomous process that is referred to as cell outage management (COM) [7]– [12]. COM can be further subdivided into COD and cell outage compensation (COC). COD aims to autonomously detect outage cells, i.e., cells that are not properly operating due to possible failures, e.g., external failure such as power supply or network connectivity, or even misconfiguration [4]– [6], [13]. On the other hand, COC refers to the automatic mitigation of the degradation effect of the outage by appropriately adjusting suitable radio parameters, such as pilot power, antenna tilt, and azimuth of the surrounding cells. The degree of compensation is usually dictated by the operator's policies, which also specify the level of performance that must be satisfied in the outage region [9].