Affiliations: School of Computer Science and Technology, Jilin University, Changchun, P. R. China | Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education, Jilin University, Changchun, P. R. China. E-mails: [email protected], [email protected]
Note: [] Corresponding author: Dantong Ouyang, School of Computer Science and Technology, Jilin University, Changchun 130012, P. R. China. Tel.: +86 138 4300 6163; E-mail: [email protected].
Abstract: Recently, model-based diagnosis of discrete event systems has attracted more and more attention. Incremental diagnosis is essential for on-line diagnosis and the diagnosis is usually performed on-line. However, the observations are often uncertain. To address this problem, a new concept of two restricted successive temporal windows is proposed. Thereafter, an approach is given to support on-line incremental model-based diagnosis of discrete event systems with uncertain observations. All the observation sequences emitted by the previous window can be produced if the second temporal window is long enough (bigger than the maximal delay of transmission). In this way, the global emitted observation sequences can be inferred as well. The proposed approach is sound, complete, timely and universal. In particular, it is well suited for on-line diagnosis of discrete event systems when the received observations are too dense to find, in time, so-called “sound windows” by traditional approaches.
