We provide EU funding and world-class technical support to engineering and technology businesses.

Find out more about who we are and what we do

Stay in Touch

EventBrite logo Twitter logo RSS logo Linked In  logo Email us

Join our mailing list

Join our mailing list to receive notification of CPSE Labs events and upcoming calls

Smart Anything Everywhere

Cyber-Physical Systems Engineering Labs is part of the Smart Anything Everywhere initiative.

This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 644400.

Integration of Co-Simulation Methods into a real-time Platform for CPS

Problem and solution

In principle, the integrated verification of a CPS can only be done once the final product is assembled: beforehand, only parts of it are available (ECU, chassis, brakes), in diverse locations (different suppliers), in different states (idea, virtual model, prototype, product), which thus prevents integrated verification. However if all the parts are virtually available via models that can be simulated, then a virtual integration can be performed. The behaviour of this virtual integration can then be verified using cosimulation. However once one of the sub-parts reaches a mature stage, its (virtual) model should be replaced by its real product counterpart. Cosimulation cannot however deal directly with real-time parts: we should actually then switch from cosimulation to HIL (Hardware in the Loop) simulation.

The problem is that cosimulation and HIL simulation are very different at the conceptual and technological level: conceptually, cosimulation has as much time as it needs to compute the outcome of a simulation, HIL simulation is constrained by real, physical time. Technologically, cosimulation can be achieved on one machine using only local communication, HIL simulation requires connecting a computer to, say, a testbed, typically via some CAN bus. As a consequence, the time to switch from virtual cosimulation to HIL simulation is very large (and it can take months just to set up the testbed).

Even though the conceptual differences between cosimulation and HIL can hardly be overcome (or if so, these are so high that overcoming them should be the subject of fundamental research and not of an experiment), intuition says that the technological ones could be made easier. After all, the architecture remains the same between the models and between the parts of the system. It should therefore be easy to adapt the technologies to facilitate the transition. One could have just a switch on the channels between the subsystems: the switch can map a channel to another virtual model during cosimulation, or to a CAN port in case of HIL simulation.

This is in a nutshell the focus of the RT-CoSim experiment. To achieve this, a realtime-platform by AVL was extended to allow the integration of simulation tools (non-realtime or realtime) and test-benches (realtime only). An additional intermediate case is also targeted: having a cosimulation where some models are simulated on a non-realtime operating system (Windows) while some others are simulated on a realtime OS. This is an additional transition step between cosimulation and HIL simulation.

The extension of AVL technology has been evaluated on a demonstrator, which was built in two phases: first, an "office-only" co-simulation was implemented on a (non-realtime Windows) PC (using AVL's Model.CONNECT). In the next phase, the same architecture was implemented replacing the engine model by a wrapper connected to a real testbed (more precisely connected to an intermediate realtime OS PC, itself connected to the testbed). This enabled the testing engineers to use the same setting for the simulation and for the real testbed.

TEMPO Simulator

The technical work of the experiment comprised the integration of the various pre-existing components of AVL and the Virtual Vehicle research centre, and the development of the necessary components ("coupling element") to connect the non-realtime with the realtime environment including sophisticated methods for error correction ("NEPCE").

How did CPSE Labs Help?

Out of this experiment, AVL and Virtual Vehicle could achieve the following benefits:

  • Extend their current software to target the above-mentioned problem.
  • Evaluate this extension on some testbed.

After various intermediate corrections, the evaluation confirmed that the extension provides the expected benefits.


As a result, AVL can now provide this technology as a part of their software and provide a clear added value to their customers: it indeed provides a solution to a well known problem. The usage of the solution itself shall speed up the development of CPS (especially in the automotive domain since it is the main market of AVL) thus reducing time to market.


Most outputs of the experiment result in extensions of the Model.Connect software of AVL. In the Design Centre, further small-scale experiments were carried out using our internal model-based software AutoFOCUS3, which, for this purpose, was extended with FMI (Functional Mockup Interface – standard for cosimulation) support.

Lessons learnt on the domain problems have been reported in "mini-courses" at the Design Centre. Introductions to cosimulation and to the addressed problem have been carried out in the context of workshops in other domains (especially in avionics via the ASSET project funded by the German Federal Ministry of Economic Affairs and Energy, which includes various avionics OEMs and suppliers, among others SMEs).

Design centre

This experiment is supported by our Germany South design centre

Germany South design centre

Technology platforms



AVL Logo
Virtual Vehicle Logo


1st Jan 2016 - 30th Jun 2017
Funded under: CPSE Labs Call 1