No abstract

This paper describes a formal method for specifying the observable (external) behavior of information systems using a process algebra and input-output traces. Its notation is mainly based on the entity concept, borrowed from the Jackson System Development method, and integrated with the requirements class diagram to represent data structures and associations. The specification process promotes modular and incremental description of the behavior of each entity through process abstraction, entity type patterns, and entity attribute function patterns. Valid system input traces result from the composition of entity traces by using parallel composition operations. The association between input traces and outputs through an input-output relation completes the specification process.

Abstract-This paper describes a synthesis method that automatically derives controllers for timed discrete-event systems with nonterminating behavior modeled by timed transition graphs and specifications of control requirements expressed by metric temporal logic (MTL) formulas. Synthesis is performed by using 1) a forward-chaining search that evaluates the satisfiability of MTL formulas over sequences of states generated by occurrences of actions and 2) a control-directed backtracking technique that takes into consideration the controllability of actions. This method has several interesting features. First, the issues of controllability, safety, liveness, and real time are integrated in a single framework. Second, the synthesis process does not require explicit storage of an entire transition structure over which formulas are checked and can be stopped at any moment, giving an approximate but useful result. Third, search and control mechanisms allow circumvention of the state explosion problem.

This paper investigates the control of parameterized discrete event systems when specifications are given in terms of predicates and satisfy a similarity assumption. This study is motivated by a weakness in current synthesis methods that do not scale well to huge systems. For systems consisting of similar processes under total or partial observation, conditions are given to deduce properties of a system of n processes (arbitrary size) from properties of a system of n 0 processes (bounded size), with n ≥ n 0 . Furthermore, it is shown how to infer a control policy for the former from the latter's, while taking into account interconnections between processes.

No abstract