Interaction Between Heterogeneous Models: A Formal Approach

Roberto Passerone and Alessandro Pinto
(Professor Alberto L. Sangiovanni-Vincentelli)
GSRC

Embedded systems often exhibit behaviors that are best described using different formalisms and models of computation. While each model offers specialized and efficient techniques to solve particular problems in their domain of application, the interaction across models is made more difficult by their differences. As a result, this interaction is not yet well understood and is often not formally defined. In this project we seek to determine a consistent way to define interactions across different models.

We regard a system as the composition of computational domains. Each computational domain is an island in which the semantics of the coordination and communication among agents is defined, and differs from the semantics of other domains. This definition is generic and covers many possible applications: from communication between models of computation as mathematical domains, to co-simulation of applications described in different programming languages.

We use the framework of trace algebra to describe the semantic domain of each computational domain. We then investigate ways to compose domains in order to obtain additional hybrid domains that exhibit properties from their components. Constructions of this kind include products of algebras and their disjoint sums. We then study the composition of heterogenous agents (agents drawn from different computational domains) in the context of a hybrid domain, and define identity elements with respect to the operation of composition. The objective of this research is to derive correct conversion mechansisms, or communicators, from the identities in the hybrid domain.

[1]
J. Burch, R. Passerone, and A. Sangiovanni-Vincentelli, "Overcoming Heterophobia: Modeling Concurrency in Heterogeneous Systems," Proc. Application of Concurrency to System Design, Newcastle, UK, June 2001.

Send mail to the author : (apinto@eecs.berkeley.edu)


Edit this abstract