An Extended Survey and a Comparison Framework for Dataflow Models of Computation and Communication
Guillaume Roumage, Selma Azaiez, Cyril Faure, Stéphane Louise
TL;DR
The paper surveys dataflow models of computation and communication (DF MoCCs) for cyber-physical systems, focusing on expressiveness vs. analyzability and proposing a quantitative framework to compare models. It introduces a comprehensive feature set, a range-rate taxonomy, and rate/topology update classifications, then maps 11 static analyses and their dependencies. The eight-model-category classification (including Synchronous, Phased-based, Timed-based, Boolean-based, Scenario-based, Meta-models, Enable/Invoke, and Process-network DF MoCCs) is used to quantify expressiveness and analyzability via a scoring protocol, demonstrated through a visualization tool. The framework supports extensibility and aims to guide designers in selecting models that balance expressiveness with tractable verification, highlighting a trade-off that underpins much of the DF MoCC literature. The authors invite community participation to extend features, analyses, and models within an open repository.
Abstract
Dataflow Model of Computation and Communications (DF MoCCs) is a formalism used to specify the behavior of Cyber-Physical Systems (CPSs). DF MoCCs are widely used in the design of CPSs, as they provide a high-level of abstraction to specify the system's behavior. DF MoCCs rules give semantics to a dataflow specification of a CPS, and static analysis algorithms rely on these semantics to guarantee safety properties of the dataflow specification, such as bounded memory usage and deadlock freeness. A wide range of DF MoCCs exists, each with its own characteristics and static analyses. This paper presents a survey of those DF MoCCs and a classification in eight categories. In addition, DF MoCCs are characterized by a comprehensive list of features and static analyses, which reflect their expressiveness and analyzability. Based on this characterization, a framework is proposed to compare the expressiveness and the analyzability of DF MoCCs quantitatively.