Ecos

Ecos (Easy co-simulation) is a fast, efficient and very easy to use co-simulation engine written in modern C++.

Ecos provides the following features:

• FMI for Co-simulation version 1.0 & 2.0.
• Partial support for FMI 3.0.
• SSP version 1.0.
• Optional per process/remote model execution.
• Post-simulation plotting using matplotlib.
• Command-line-interface (CLI).
• Simplified Python and C interface.
• Minimal (and automatically resolved) build dependencies.

Who is this for?

Ecos is a framework for performing FMI-based co-simulation. Several such libraries/tools exist and cover much of the same ground. However, the value-proposition of Ecos is the "ease-of-deployment", while still being efficient. The library is written in Object-Oriented C++, supports all versions of FMI for Co-simulation and provides a pypi hosted Python interface. The framework can be used either in C++, C, Python or through a CLI. Moreover, it provides seamless integration with plotting (for quick prototyping) and per-process execution of model instances (remoting). The project is maintained as a mono-repo with small and few dependencies, making building from a source a breeze. This is especially valuable in an educational setting, where emphasis should be on how to use and not how to build.

Building

Ecos bundles all required dependencies. Just build.

//windows
cmake . -A x64 -B build -DECOS_BUILD_EXAMPLES=ON -DECOS_BUILD_TESTS=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build --config Release

//linux
cmake . -B build -DECOS_BUILD_EXAMPLES=ON -DECOS_BUILD_TESTS=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build

Consuming C/C++ library

Ecos is compatible with CMake FetchContent:

include(FetchContent)
set(ECOS_BUILD_CLI OFF)     # Set to ON for building ecos command-line-interface
set(ECOS_BUILD_CLIB OFF)    # Set to ON for building C API
set(ECOS_WITH_PROXYFMU OFF) # Set to ON for remoting
FetchContent_Declare(
        ecos
        GIT_REPOSITORY https://github.com/Ecos-platform/ecos.git
        GIT_TAG git_tag_or_commit_id
)
FetchContent_MakeAvailable(ecos)

add_executable(ecos_standalone main.cpp)
target_link_libraries(ecos_standalone PRIVATE libecos) # or libecosc for C API

Features

Per process / remote execution

Ecos enables models to run on separate processes, possibly on another PC.
Simply prepend proxyfmu://localhost?file= to the path of the fmu(s) you load.
When targeting an external proxyfmu instance, replace localhost with host:port. For each FMU instance created when using proxyfmu a new process is created. When targeting localhost, Unix Domain Sockets are used, while TCP/IP is used when targeting a remote process started with proxufmu boot --port <portNumber>. To successfully make use of proxyfmu, make sure the executable built by the project is located in the working directory of your application or added to PATH when targeting localhost. Using the Python API, however, this should work out-of-the-box. proxyfmu is implemented using simplesocket in conjunction with flexbuffers.

Ecos may be built without this feature (less dependencies, faster build) by passing -DECOS_WITH_PROXYFMU=OFF to CMake.

Scenario configuration

Scenarios in Ecos are actions to be performed during the simulation. Scenarios are most useful in a CLI context, and can be specified using the --scenarioConfig option. The structure of a scenario file follows the XML schema defined in resources/scema/ScenarioConfig.xsd. Loading a scenario through the API is demonstrated in here

CSV logging

Ecos supports CSV logging of simulation data. Which variables to log, and how often, is configurable. In a CLI context, this is done using the --csvConfig option with a path to an XML configuration file adhering to the CsvConfig.xsd schema located in resources/schema/. The API also provides the means to configure this programmatically. See /examples for various demonstrations.

Plotting

Ecos supports out-of-the-box plotting of simulation data using matplotlib in both C++ and Python. Time- and XY series plots can be configured inline or using the ChartConfig.xsd XML schema located in resources/schema/. See /examples for demonstrations.

C++ Example

using namespace ecos;

int main() {
    
    simulation_structure ss;

    // add models
    ss.add_model("chassis", "chassis.fmu");
    ss.add_model("ground", "ground.fmu");
    ss.add_model("wheel", "wheel.fmu");
    
    //make connections
    ss.make_connection<double>("chassis::p.e", "wheel::p1.e");
    ss.make_connection<double>("wheel::p1.f", "chassis::p.f");
    ss.make_connection<double>("wheel::p.e", "ground::p.e");
    ss.make_connection<double>("ground::p.f", "wheel::p.f");
    
    // setup initialValues
    std::map<variable_identifier, scalar_value> map;
    map["chassis::C.mChassis"] = 4000.0;
    ss.add_parameter_set("initialValues", map);
    
    double stepSize = 1.0 / 100; // 100 Hz
    auto sim = ss.load(std::make_unique<fixed_step_algorithm>(stepSize));
    
    // setup csv logging
    csv_config config;
    config.register_variable("chassis::*"); // logs all chassis variables
    
    auto csvWriter = std::make_unique<csv_writer>("results.csv", config);
    const auto outputPath = csvWriter->output_path();
    sim->add_listener("writer", std::move(csvWriter));
    
    // apply named set of parameters
    sim->init("initialValues");
    sim->step_until(10);
    
    sim->terminate();
    
    plot_csv(outputPath, "ChartConfig.xml"); // plot results
}

C++ SSP example

using namespace ecos;

int main() {
    
    auto ss = load_ssp("quarter-truck.ssp"); // accepts .ssp file or extracted folder

    double stepSize = 1.0 / 100; // 100 Hz
    auto sim = ss->load(std::make_unique<fixed_step_algorithm>(stepSize));
    
    // ... see example above for further usage
}

Command line interface (CLI)

Options:
  -h,--help                   Print this help message and exit
  -v,--version                Display program version information and exit
  -i,--interactive            Make execution interactive.
  --noCsv                     Disable CSV logging.
  --noParallel                Run single-threaded.
  --path REQUIRED             Location of the fmu/ssp to simulate.
  --stopTime [1]              Simulation end.
  --startTime [0]             Simulation start.
  --stepSize REQUIRED         Simulation stepSize.
  --rtf [-1]                  Target real time factor (non-positive number -> inf).
  --csvConfig                 Path to CSV configuration.
  --chartConfig               Path to chart configuration.
  --scenarioConfig            Path to scenario configuration.
  -l,--logLevel ENUM:value in {trace->0,debug->1,info->2,warn->3,err->4,off->5} OR {0,1,2,3,4,5}
                              Specify log level.

Python interface

pip install ecospy

Local pip installation

To install the python package locally:

1. Clone the project
2. Run CMake installation (see Building)
3. Run pip install .

Note: You will need wheel in addition to the compile-time requirements listed further below.

If using an old pip version, append --use-feature=in-tree-build if you get an error about ../version.txt

Python Example

print(f"Ecoslib version: {EcosLib.version()}")

EcosLib.set_log_level("debug")

fmu_path = "BouncingBall.fmu"
result_file = f"results/bouncing_ball.csv"

with EcosSimulationStructure() as ss:
  ss.add_model("ball", fmu_path)
  
  with(EcosSimulation(structure=ss, step_size=1/100)) as sim:
  
      sim.add_csv_writer(result_file)
      sim.init()
      sim.step_until(10)
      sim.terminate()

config = TimeSeriesConfig(
    title="BouncingBall",
    y_label="Height[m]",
    identifiers=["ball::h"])
plotter = Plotter(result_file, config)
plotter.show()

Colab Notebook Example (pythonfmu + ecospy)

Mass-spring-damper system

---

Compile-time requirements

• Windows (10 >=) or Ubuntu (20.04 >=)
• C++20 compiler (MSVC >= 17 || gcc11 >=)
• CMake >= 3.19

Additional Linux requirements

sudo apt install libtbb-dev unzip

Run-time requirements

• Python3 (optional, required for plotting)
  • matplotlib
  • pandas

Acknowledgments

This software is made possible thanks to the following third-party projects:

• CLI11
• flatbuffers
• pugixml
• spdlog
• fmi4c
• subprocess.h
• Catch2

---

Want to build FMUs in C++? Check out FMU4cpp
Want to build FMUs in Python? Check out PythonFMU