Solving Diagnosability of Hybrid Systems via Abstraction and Discrete Event Techniques

This paper addresses the problem of determining the diagnosability of hybrid systems by abstracting hybrid models to a discrete event setting. From the continuous model the abstraction only remembers two pieces of information: indiscernability between modes (when they are guaranteed to generate diff...

ver descrição completa

Detalhes bibliográficos
Autores: Grastien, Alban, Travé-Massuyès, Louise, Puig Cayuela, Vicenç|||0000-0002-6364-6429
Tipo de documento: artigo
Data de publicação:2017
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositório:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglês
OAI Identifier:oai:upcommons.upc.edu:2117/110149
Acesso em linha:https://hdl.handle.net/2117/110149
https://dx.doi.org/10.1016/j.ifacol.2017.08.911
Access Level:Acceso aberto
Palavra-chave:Discrete-time systems
Diagnosability
discrete event systems
Hybrid systems
Invariant sets
Sistemes de temps discret
Anàlisi de sistemes
Àrees temàtiques de la UPC::Informàtica
Descrição
Resumo:This paper addresses the problem of determining the diagnosability of hybrid systems by abstracting hybrid models to a discrete event setting. From the continuous model the abstraction only remembers two pieces of information: indiscernability between modes (when they are guaranteed to generate different observations) and ephemerality (when the system cannot stay forever in a given set of modes). Then, we use standard discrete event system diagnosability algorithms. The second contribution is an iterative approach to diagnosability that starts from the most abstract discrete event model of the hybrid system. If it is diagnosable, that means that the hybrid system is diagnosable. If it is not, the counterexample generated by the diagnosability procedure is analysed to refine the DES. If no refinement is found, then it can not be proved that the hybrid system is diagnosable. Otherwise, the refinement is included in the abstract DES model and the diagnosability procedure continues.