Options
2012
Master Thesis
Titel
Unsupervised clustering of automotive network traces for learning deterministic finite state machines
Abstract
In todays automobiles a huge amount of functionality is realized by numerous distributed embedded systems. Writing decentralized and safe software components for such a system is a highly challenging task. The test cases for these components are often generated from specification, which can be incomplete or wrong. Therefore, a model based test approach was suggested by Fraunhofer ESK. A generated dependency model from observations of a car includes also behavior, which was probably not considered in the specification. This thesis investigates, how the inference of this dependency model based on an existing implementation of an inference algorithm can be improved. On the one hand, the complexity of the inference of the model is decreased in special cases using an unsupervised clustering approach combined with evolutionary strategies. On the other hand, quality parameter of the inferred models are improved.
ThesisNote
München, Hochschule, Master Thesis, 2012
Author(s)
Advisor
Verlagsort
München
Language
English