DSML4CS: An Executable Domain-Specific Modeling Language for Co-Simulation Service in CPS

Introduction

Cyber-physical system (CPS) (Alur, 2015; Baheti & Gill, 2011) is a kind of complex system in which the computational and physical processes interact tightly in the network environment, and there is a causal logic relationship between subsystems. The CPS achieves real-time sensing and dynamic control and provides information regarding engineering system services, which makes the system more reliable, efficient and collaborative in real-time. Nowadays, typical applications for CPS cover a wide range of areas, including autonomous vehicles, remote precision surgical systems, smart grid and smart buildings, etc. CPS has three main characteristics: (1) hybrid system: It highly integrates computational and physical components; (2) safety-critical application: Computing devices control aircrafts, automobiles, and medical devices and thus are all examples of safety-critical cyber physical systems. In this context, establishing that the system works correctly at design time is of paramount importance. (3) heterogeneous: It tightly couples computation, communication, and control along with physical dynamics, which are traditionally considered separately.

Design of CPS requires the integrated use of domain-specific modeling and analysis methods that have been developed over the past decades in disparate areas of computing and networking as well as physical systems engineering such as mechanical, thermal, electrical, electronic, hydraulic, and other domains (Neema et al., 2019). CPS involves both continuous behavior of physical components and discrete behavior of computing components. Computing resources, physical equipment and subsystems of the external environment have completely different domain knowledge backgrounds, and they also have their own design and modeling platforms. Therefore, it is still a challenge problem to conduct the unified design modeling and simulation of all the above-mentioned subsystems with the Domain Specific Modeling Language (DSML) through an integrated platform.

Co-simulation service is a promising technology to solve these problems, which allows modeling tools to operate with their respective solver (to protect their intellectual property) and co-simulate these models with a unified interface. Thus, multiple heterogeneous models can be coordinated and communicated to perform system functions. This feature of co-simulation service solves the common problem of the single model simulation and provides a novel way to simulate the behavior of CPS. However, it is hard to model the coordination between heterogeneous subsystems.

In this work, the authors’ aim is to solve the problems of modeling and simulating heterogeneous CPS and to better express the coordination and connection between subsystems. The authors propose the Domain Specific Modeling Language for Co-Simulation (DSML4CS) to model the co-simulation between heterogeneous FMUs. DSML4CS has been implemented based on the open source modeling framework GEMOC Studio. It supports modeling co-simulation of CPS with DSML4CS and provides co-simulation services for simulating the hybrid behavior of CPS.

In short, our main contributions are as follows:

• •

  A co-simulation modeling language DSML4CS for heterogeneous CPS is proposed to facilitate the modeling of co-simulation, which is based on the meta-modeling approach and domain specific language. The syntax and semantics are presented based on the domain knowledges of CPS;

• •

  The technical framework for DSML4CS is presented to enable the tool implementation based on GEMOC studio. The abstract syntax of DSML4CS is implemented based on EMF, the concrete syntax is implemented based on Sirius, and operation semantics is implemented based on Xtend;

• •

  The authors model a specific scenario of the temperature control system in an energy-aware building based on DSML4CS tool. The authors explore simulation experiments with three different coordinators: fixed step size, variable step size, and coordinator with partial step. The experimental results demonstrate the usability and feasibility of the proposed approach.