Cooperative Localization for Networks with Dynamic Connectivity
Author
Voulgaris, Konstantinos
Term
4. term
Education
Publication year
2016
Submitted on
2016-06-01
Pages
107
Abstract
Dette speciale undersøger forbindelsesdynamikken i kooperativ lokalisering, hvor enheder forbedrer deres positionsestimat ved at dele information. Arbejdet modellerer netværket som en graf og udvider en tidsuafhængig probabilistisk konnektivitetsmodel til tidsdomænet ved at konstruere en Markov-kæde, hvis stationære fordeling svarer til den oprindelige model. For at beskrive den tidslige udvikling analyseres autokorrelationsfunktioner, og der introduceres et mål for grafens konsistens over tid baseret på Jaccard-indekset. Specialet omfatter også en mobilitetsbeskrivelse, en rækkefejlmodel og kalibrering af modelparametre via maksimum likelihood og maksimum a posteriori estimering, herunder samlet estimering af nøgleparametre. Metoden evalueres gennem simuleringer og sammenlignes med etablerede konnektivitetsmodeller i litteraturen (bl.a. Savic og Zazo samt Henk Wymeersch). Resultaterne viser, at den foreslåede Markov-model kan indfange de karakteristika, som den eksisterende model udtrykker, og at den udviklede estimeringsstrategi er velegnet til modelkalibrering.
This thesis studies connectivity dynamics in cooperative localization, where devices improve position estimates by sharing information. The network is represented as a graph, and a time-independent probabilistic connectivity model is extended into the time domain by constructing a Markov chain whose stationary distribution matches the original model. To capture temporal behavior, the autocorrelation function is analyzed, and a metric for graph consistency over time based on the Jaccard index is introduced. The work includes a mobility description, a range error model, and parameter calibration using maximum likelihood and maximum a posteriori estimation, including joint estimation of key parameters. The approach is evaluated via simulations and compared with established connectivity models in the literature (including those by Savic and Zazo and by Henk Wymeersch). Findings indicate that the proposed Markov model reproduces the characteristics of the existing model and that the developed estimation strategy is suitable for model calibration.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents