Skip to Main content Skip to Navigation
Theses

Formal methods for distributed real-time systems

Résumé : Nowadays, real-time systems are ubiquitous in several application domains.Such an emergence led to an increasing need of performance (resources,availability, concurrency, etc.) and initiated a shift from theuse of single processor based hardware platforms, to large setsof interconnected and distributed computing nodes. This trend introduced the birthof a new family of systems that are intrinsically distributed, namelyemph{Networked Embedded Systems}.Such an evolution stems from the growing complexity of real-time softwareembedded on such platforms (e.g. electronic control in avionicsand automotive domains), and the need to integrate formerly isolated systems so thatthey can cooperate, as well as share resources improving thus functionalitiesand reducing costs.Undoubtedly, the design, implementation and verification of such systems areacknowledged to be very hard tasks since theyare prone to different kinds of factors, such as communication delays, CPU(s)speed or even hardware imprecisions, which increases considerably the complexity ofcoordinating parallel activities.In this thesis, we propose a rigorous design flow intended forbuilding distributed real-time applications.We investigate timed automata based models, with formally defined semantics, in orderto study the behavior of a given system with some imposed timing constraints when deployedin a distributed environment. Particularly, we study emph{(i)} the impact of the communicationdelays by introducing a minimum latency between actions executions and the effectivedate at which actions executions have been decided,and emph{(ii)} the effect of hardware imperfections, more precisely clocks imprecisions,on systems execution by breaking the perfect clocks hypothesis, often adopted duringthe modeling phase. Nevertheless, timed automata formalism is intended to describe a highlevel abstraction of the behavior of a given application.Therefore, we use an intermediate representation ofthe initial application that, besides having say{equivalent} behavior, explicitly expressesimplementation mechanisms, and thus reduces the gap between the modeling and the concreteimplementation. Additionally, we contribute in building such systems by emph{(iii)}proposing a knowledge based optimization method that aims to eliminate unnecessarycomputation time or exchange of messages during the execution.We compare the behavior of each proposed model to the initial high level model and study therelationships between both. Then, we identify and formally characterize the potential problemsresulting from these additional constraints. Furthermore, we propose execution strategies thatallow to preserve some desired properties and reach a say{similar} execution scenario,faithful to the original specifications.
Document type :
Theses
Complete list of metadatas

Cited literature [99 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02146390
Contributor : Abes Star :  Contact
Submitted on : Monday, June 3, 2019 - 9:19:14 PM
Last modification on : Friday, August 7, 2020 - 3:01:02 AM

File

DELLABANI_2018_archivage.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02146390, version 1

Collections

Citation

Mahieddine Dellabani. Formal methods for distributed real-time systems. Embedded Systems. Université Grenoble Alpes, 2018. English. ⟨NNT : 2018GREAM090⟩. ⟨tel-02146390⟩

Share

Metrics

Record views

196

Files downloads

87