Translating a large subset of stateflow to hybrid CSP with code optimization

Abstract

Stateflow is a graphical language for modeling hierarchical transition systems, well-known for the complexity of its semantics, which is only informally explained in its user manual. Formal analysis and verification of Stateflow models usually proceed by first translating a subset of Stateflow to a formal language with precise semantics. Most existing work address only “safe” subset of Stateflow and ignore the most complex semantic issues. Moreover, it is difficult to balance simplicity of the translation algorithm with conciseness of the resulting model. In this paper, we describe a two-stage process for translating a large subset of Stateflow to Hybrid CSP, where the first stage is mostly syntax-directed and addresses each feature of Stateflow separately, and the second stage is a code optimization step that simplifies the resulting model using information from static analysis. We further incorporate this method into translation of Simulink models, in particular consider Stateflow charts triggered by various signal sources and function calls. The translation process is thoroughly validated using a hand-designed set of benchmarks, as well as larger case studies from existing work.