Transaction-level Functional Mockup Units for Cyber-Physical Virtual Platforms

Abstract

The modelling process of Cyber-physical Systems aggregates semantics and languages tailored to different specific domains. The simulation of these complex systems involves different tools and their coupling requires computational effort. In the last few years both Academy and Industry worked toward the definition of standard interfaces able to overcome such issues. The Functional Mock-up Interface (FMI) standard emerged as one of the most promising tool to easily export and integrate heterogeneous models. However, the standard still shows some weaknesses, particularly when dealing with Functional Mock-up Units (FMUs) describing discrete-event systems. This paper explore the features of the standard to find its shortcomings when dealing with discrete models. Then, it proposes a systematic approach to fully exploit the features of the current standards to overcome such limitations. The solution is based on two concepts: (1) exposing the internal time of the FMUs, and (2) exploits the newly exposed information to implement temporal decoupling. The combination of these two concepts allows to optimize the FMUs coordination algorithms. It reduces the number synchronization points and move the simulation from cycle-accurate to transaction-accurate. The impact of these optimizations is measured on a set of benchmarks having different tread-offs of computation and control.