1. FMI & Co-simulation Bert Van Acker, Cláudio Gomes and Joachim Denil
Good afternoon,
I’m Claudio Gomes and I’m here to present you our work in Co-simulation.
Everyone in the group contributes to this work but the lead authors are Bert, me and Joachim.

2. The Modern Car Complexity Competitive Market Concurrent Development
40+ subsystems
Competitive Market
Concurrent Development
Late Integration Nightmare
Distributed Development
OEM suppliers
Late Integration (due to IP)
The modern car is has more than 40 subsystems.
And, in a competitive market such as the one we live in, industries cannot afford to develop all these subsystems sequentially.
This means that inevitably they have to integrate all these systems.
The more late they integrate them, the worse.
To aggravate the situation, these subsystems are not developed in house: there are specialized independent suppliers that develop these subsystems.
This means the company that is making cars, is an integrator. It receives these OEM components, frequently as black boxes to protect the supplier’s intellectual property,
And they have to coupled them together. *

3. Functional Mock-up Interface (FMI)
Xml + C Representation for Models
Standard
Tool Independent
Black box*
Earlier integration of OEM
IP Protection (dll)
Reduced cost & TTM
(Co-)Simulation
This is were the functional mock up interface standard comes into play.
It is a standard to represent models (which can be systems of equations, state machines, etc…) using C code and xml files.
The C code contains the code necessary to interact with the model. For instance, if the model is a system of equations, then the c code will allow you to compute the values of output variables, given the values of input variables.
The xml file contains metadata about the input and output variables, their types, etc…
This makes the standard tool independent. Any modelling tool can export a model into C.
And because the model is represented in C code, it is black box.
This standard aims to allow earlier integration of OEM components. So the supplier can give the company a dll and an xml file, and the company can interact with those dlls before buying the hardware.
This will reduce the integration problem and the time to marked.
But off course, one does not simply interact with the supplied dlls individually. The company must couple those dlls with its own models and perform co-simulation. *

4. Co-Simulation Coordination HiL Simulation Hybrid Simulation*
Coordination
This is an ABS system.
As tou can see, there are at least 4 kinds of components:
The control unit;
The Hydraulic modulator;
The wheel sensor;
And the wheel brake.
With FMI the company can get the FMUs for each component and couple them like this and perform co-simulation to check if the components work well together.
If you look closely to the individual dlls and xmls, or Functional Mockup Units, you will see that they can be comprised of a Model and a Solver.
The model contains the equations, or the state machine, and the solver knows how to simulate the equations or the state machine.
But off course, these solvers files don’t know each other.
So we cannot just couple them together and hope that they simulate in a synchronized fashion.
We need a coordinator that commands each solver individually to achieve a correct co-simulation of all FMUs.
That’s the purpose of the master algorithm. It takes outputs of FMUs, plugs them in in other FMUs, performs simulation steps to get new outputs, and so on.
This is challenging because we can have discrete FMUs (such as state machines) communicating with continuous FMUs such as systems of equations and so, we need to perform semantic adaptation of these formalisms.
Another challenge, is to enable Hardware in the loop simulation, by wrapping a component as an FMU and plugging it into the co-simulation.
Hybrid Simulation
Semantic Adaptation
HiL Simulation

5. Specific Master Generation
Efficiency
Correct co-simulation
To generate the Master, our approach is to:
Take the set of FMUs;
Take the co-simulation model which is the coupling of the FMUs, perform some extra analysis to understand how the FMUs are connected and their nature,
And then generate a master algorithm which is as simple and as specific as possible for that co-simulation model.
This approach is the basis for achieving simplicity of the master algorithm, which means efficiency and transparency because then we can model the master algorithm.
And at that point, we can perform model checking to prove properties about the master.

6. References
Van Acker, B., Denil, J., Vangheluwe, H., De Meulenaere, P., “Generation of an Optimised Master Algorithm for FMI Co-Simulation”, in Proceedings of the Symposium on Theory of Modeling & Simulation-DEVS Integrative, Society for Computer Simulation International (2015), p
Bart Pussig, Joachim Denil, Paul De Meulenaere, Generation of Co-simulation Compliant Functional Mock-up Units From Simulink, Using Explicit Computational Semantics, Proceedings of the 2014 Symposium on Theory of Modeling & Simulation: DEVS Integrative M&S Symposium, 2014
Joachim Denil, Bart Meyers, Paul De Meuleneare, Hans Vangheluwe: Explicit Semantic Adaptation of Hybrid Formalisms for FMI Co-Simulation, TMS-DEVS, 2015