文章基本信息

标题：Comparing Linear Systems with Gaussian Mixture Model Additive Uncertainties Using Kullback-Leibler Rate
本地全文：下载
作者：Aditya Karumanchi ; Punit Tulpule
期刊名称：IFAC PapersOnLine
印刷版ISSN：2405-8963
出版年度：2021
卷号：54
期号：20
页码：566-572
DOI：10.1016/j.ifacol.2021.11.232
语种：English
出版社：Elsevier
摘要：AbstractIn simulation-based testing of complex systems, it is essential to deal with multiple fidelities of subsystem models, which necessitates comparison of such models. While models with Gaussian additive uncertainties can be compared easily, comparison of models with more general probability density functions (pdfs) is more challenging. In this study, we examine the propagation of additive uncertainty with a general pdf, modeled using Gaussian Mixture Models (GMMs), and subsequent comparison of model uncertainties between two models. The proposed approach is based on Kullback Leibler (K-L) rate pseudo metric. We also propose a method based on K-L divergence for pruning the number of GMM components as they are propagated through a linear system, in order to reduce computational cost. We exemplify the proposed approach using two models of a vehicle platoon.
关键词：KeywordsGaussian Mixture ModelsModel uncertainty comparisonNon-Gaussian uncertaintyMonte Carlo approximationKullback-Leibler divergenceK-L divergence approximation