Skip to Main content Skip to Navigation

Exploration concrétisée et pertinente de systèmes d'événements abstraits en vue de la génération automatique de tests

Abstract : The work presented in this thesis contributes to the automated under-approximation generation techniques in order to generate tests from models.Model based testing aims at guaranteeing the conformity of an implementation with regards to a model, both designed from specifications by two distinct teams.In this thesis, we propose to use well known abstraction techniques by using predicates extracted from behavioural models allowing to reduce the manipulated state space to a finite and narrow set. We propose to extract the abstraction predicates from the test purpose so that the generated tests cover the behaviours it targets. However, the computation of an abstraction causes a loss of information of the reachability with regards to the initial model. Our goals are, on the one hand, to efficiently compute an abstraction of the model as representative as possible of a test purpose. On the other hand, we strive to extract executions from this abstraction which can be instanciated on the model before its abstraction, targeting the behaviours to test, and aiming at covering the states and the transitions of the abstracted model.Our contributions are the following. We define a test generation method combining various algorithms allowing to obtain a good structural coverage of the abstraction of a non deterministic behavioural model. We first propose an algorithm computing the abstraction of a model using predicates extracted from a test purpose expressed as a temporal property. This algorithm computes an under-approximation of the model by covering the abstract states and transitions of the model. It applies various exploration heuristics and techniques with the aim of increasing the amount of reached instances. Then, we propose to improve by the means of two other algorithms the structural coverage obtained by this first under-approximation. The first of these algorithms, fully automated, takes advantage of the known modalities of the abstract transitions which provide us with reachability properties. The second under-approximation extension algorithm uses the expertise of the tester who must, from non covered transitions, design a relevance predicate which will guide and limit the exploration and the instanciations. We define a set of rules allowing to express this relevance predicate and compute a variant guaranteeing the termination of the exploration algorithm. These two algorithms complete the under-approximation by necessarily instanciable executions. Finally, we experimentally assess the quality of the method on five case studies.
Complete list of metadatas

Cited literature [94 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Monday, March 9, 2020 - 10:33:09 AM
Last modification on : Friday, October 23, 2020 - 5:00:13 PM
Long-term archiving on: : Wednesday, June 10, 2020 - 1:42:20 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02502304, version 1


Guillaume Voiron. Exploration concrétisée et pertinente de systèmes d'événements abstraits en vue de la génération automatique de tests. Autre [cs.OH]. Université Bourgogne Franche-Comté, 2019. Français. ⟨NNT : 2019UBFCD027⟩. ⟨tel-02502304⟩



Record views


Files downloads