LiU » ELLIIT » Scalable Language Tools for Cyber-Physical Systems

Scalable Language Tools for Cyber-Physical Systems

G. Hedin, LU
Participants at LiU: P. Fritzson, Adrian Pop
Participants at HH: W. TahaW. Mostowski

Project description: Swedish industry is among the world leading on modeling of cyber-physical systems. By using
high-level modeling languages like Modelica, complex physical systems can be modeled in a compact and natural
way, reusing libraries for different engineering domains like mechanical, electrical, thermal, control, etc. Advancing
this area is of great strategic importance. In this project, we focus on scalable techniques for performance, novel
tooling, and for extending the application area to generating control software for cyber-physical systems.

Our research builds on open-source platforms for the Modelica language: OpenModelica and, and
open-source platforms for metacompilation: RML/MetaModelica and JastAdd. These systems are extended and
applied to demonstrate results in the form of new algorithms, tools, and language constructs. The project focuses
on aspects that support scaling to large applications, in particular performance, development support, and control
software generation.

For large applications, performance becomes a bottleneck. To counter this, we will develop new techniques for
automated parallelization of compilation as well as simulation, using both data and control parallelism. The development
of large applications requires advanced tooling for editing, debugging, and verifying models. We will develop
novel tools with particular focus on interactive visual smart editing, equational debugging, and support for model
requirements as well as testing and verification of requirements. Modeling is traditionally used for simulation, but
extending their application to generate control software is an active research area. We will work on new language
constructs for supporting the generation of embedded control software directly from models, for example in the form
of FMUs (Functional Mockup Units). We will demonstrate the new results on large models in the order of 100k