Browsing by Author "Sousa, Marco Décio Baptista"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- Self-diagnosing and optimization of low coverage and high interference in 3G/4G: radio access networksPublication . Sousa, Marco Décio Baptista; Vieira, Pedro Manuel de Almeida Carvalho; Martins, André Eduardo PoncianoSelf-Organizing Networks (SON) solutions have been developed and implemented in the last years as a Mobile Network Operator (MNO) strategy to deal with the complexity of current networks. This research work, focuses on the self-optimization branch of SON solutions. It aims to empower a network with automatic capabilities for detecting and optimizing poor Radio Frequency (RF) performance scenarios. The detection and optimization of those scenarios, is based on Drive Test (DT) data. This leads to the development of a DT classi cation model to assert the quality of data collected through DT for a given cell, as it supports all decision making in terms of detection and optimization of poor RF situations. The DT model was calibrated with subjective testing in the form of inquiries made to fty Radio Access Network (RAN) engineers. Three algorithms were implemented for detection of low coverage and high interference scenarios. Besides identifying and dividing into clusters the DT data that denotes each problem, harshness metrics at cell and cluster level allow to identify the most severe situations. Moreover, an antenna physical parameter optimization algorithm, based on a Particle Swarm Optimization (PSO) algorithm, is able to purpose new Electrical Downtilt (EDT), Mechanical Downtilt (MDT) or the antenna orientation to improve or x the detected RF problems. All algorithms were tested with real MNO DT data and network topology, mainly on urban scenarios, where the detection and optimization is more critical for MNO. Regarding the detection algorithms, in urban scenario, it was established that the situations of high interference were more prevailing than the low coverage. The antenna self-optimization algorithm achieved an average gain of 78% on the tested cases.