Search

Fluid modeling is a technique for approximating behavior of stochastic discrete-state systems based on dynamical models with continuous state variables. In this paper, we focus on fluid modeling for Markovian stochastic Petri nets, and propose a method to compute regions including state values that appear with a high probability. This can be seen as the second-order approximation, because the size of each region varies according to the variance of the corresponding state variable. The fluid models are represented by discrete-time dynamical systems with interval parameters, and interval arithmetic is used for computing regions at each time step. We show simulation results on the fluid models and compare the results with the correct solutions obtained by discrete-state analysis on stochastic Petri nets. [ABSTRACT FROM AUTHOR]

We propose an algorithm for thesynthesis of supervisors in discrete event systems. The algorithmis based on a learning algorithm of regular languages proposedby Angluin, and constructs a supervisor realizing an unknownspecification which is identified through the interaction betweenthe designer and the algorithm. We also consider the synthesisproblem for systems consisting of several processes which behaveconcurrently. One of serious problems in dealing with such aconcurrent system is that the number of states required for describingthe global behavior often grows exponentially in the size ofthe model. To improve this situation, we introduce the conceptof dependency defined on the set of events. It prevents the algorithmfrom considering all interleavings of independent occurrencesof events. [ABSTRACT FROM AUTHOR]