Implementation of a self-healing framework
Author
Khatri, Vikramajeet
Term
4. term
Publication year
2011
Submitted on
2011-05-31
Pages
50
Abstract
This thesis investigates how to design and implement a self‑healing framework for LTE‑based self‑organizing networks (SON) to automate fault management, uphold SLAs, and reduce CAPEX/OPEX. It defines an LTE‑specific fault‑to‑symptom mapping and proposes an architecture that combines threshold setting, probability estimation, and diagnosis. Thresholds for key metrics (e.g., downlink SINR) are derived using entropy‑based discretization (EMD, SEMD) and a beta‑distribution‑based MAP approach, while probabilities are estimated via maximum likelihood and M‑estimate. Diagnosis is modeled using both Bayesian networks and a codebook approach that links observed symptoms to likely causes. The framework is evaluated in MATLAB through an LTE radio network simulation with user mobility, analyzing uplink/downlink interference, downlink coverage, and shadowing introduced by obstacles. Reported outputs include per‑user thresholds, counts of frames affected by interference and shadowing, cause probabilities, and detection accuracy. Specific numerical results are not present in this excerpt, but the text indicates the method can distinguish interference‑ and shadowing‑related issues and quantify detection performance. The thesis concludes with findings and directions for future work.
Denne afhandling undersøger, hvordan et selvhelende rammeværk kan designes og implementeres for LTE-baserede selvorganiserende mobilnet (SON) for at automatisere fejlstyring, opretholde SLA’er og reducere CAPEX/OPEX. Arbejdet formulerer en fejl‑til‑symptom‑kortlægning for LTE og foreslår en arkitektur, der kombinerer tærskelsætning, sandsynlighedsestimering og diagnoser. Tærskler for nøglemålinger (fx nedlink‑SINR) fastsættes med entropibaserede metoder (EMD, SEMD) og en beta‑fordelingsbaseret MAP‑tilgang, mens sandsynligheder estimeres med maksimum-likelihood og M‑estimat. Diagnosen modelleres både med Bayesianske netværk og en kodebogstilgang, der kobler observerede symptomer til sandsynlige årsager. Frameworket afprøves i MATLAB gennem en LTE‑radionetværkssimulering med brugermobilitet, hvor op‑ og nedlinkinterferens, dækning og skyggetab (via tilføjede forhindringer) analyseres. Resultaterne omfatter per‑bruger‑tærskler, antal rammer påvirket af interferens og skygning samt beregnede årsagssandsynligheder og detektionsnøjagtighed. De konkrete tal fremgår ikke af dette uddrag, men teksten indikerer, at metoden kan skelne mellem interferens‑ og skyggerelaterede problemer og kvantificere detektionsydelse. Afhandlingen afslutter med konklusioner og forslag til fremtidigt arbejde.
[This apstract has been generated with the help of AI directly from the project full text]
Keywords