MultiCoSim

Library for managing multi-fidelity co-simulations of cyber-physical systems.

Installation

This library can be installed either from PyPI using the command pip install multicosim or from github by using the command pip install https://github.com/cpslab-asu/multicosim#egg=multicosim.

This library requires multiple docker containers to be available on the system. In general, these containers will be downloaded if they do not exist, but they can also be built from source if necessary. Instructions for building these containers are in the Building From Source section.

Usage

In order to execute a simulation, import the system you are interested in simulating and provide the appropriate configuration objects to customize the simulation. The following is a simple example to demonstrate some of the configuration options available.

import multicosim as mcs
import multicosim.gazebo as gz

px4 = mcs.PX4(model=mcs.PX4.Model.X500)
gazebo = mcs.Gazebo(backend=gz.ODE(), step_size=0.001)
poses = px4.simulate(gazebo)

You can also use this library to define your own systems for execution. In order to execute a system there are two components that must be provided. The first the firmware execution program that will run in the docker container and communicate with the Gazebo simulator. This can be quickly implemented using the provided multicosim.serve decorator like so:

# controller.py

import multicosim as mcs


class Config: ...


class Result: ...


@mcs.serve()
def server(msg: Config) -> Result:
    # Execute firmware
    ...


if __name__ == "__main__":
    server()

For this component, the msg parameter will be the datastructure that is sent to the firmware to start the simulation, and the return value from the decorated function will be the datastructure that is sent back when the simulation is complete and should contain the result data of the simulation. This program needs to be loaded into a container image that is accessible by the docker context the library is executed using.

The second required component is the configuration of an executor for the newly defined system container. This can be implemented using the multicosim.manage decorator like so:

# executor.py

import multicosim as mcs

import controller

@mcs.manage(
    firmware_image=...,
    gazebo_image=...,
    command=...,
    port=...,
    rtype=controller.Result,
)
def system(world: str, x: int, y: str) -> controller.Start:
    ...

if __name__ == "__main__":
    gz = mcs.Gazebo()
    result = system.run(gz, 10, "foo")

In this example, the firmware_image argument contains the name of the container image for executing the firmware defined in the previous example, and the gazebo_image argument defines the gazebo container image to use for simulation, which might contain additional models or plugins depending on the simulation requirements. The command argument specifies what command to execute for the system, and the rtype paramter defines the type that should be sent back from the firmware when the simulation is terminated. The return value of the wrapped function is the message that will be sent to the firmware to start the simulation. The wrapped function must accept at minimum a world argument which is the name of the currently executing Gazebo world that can be used to communicate with the simulator using the transport libraries. To execute a simulation, the run method can be called, which accepts a multicosim.Gazebo instance representing the simulator configuration and all the arguments of the wrapped function following the world argument.

Building From Source

This project is the built using Hatch which is a packaging and library management tool similar to Poetry. To build this project, ensure that you have the hatch binary available somewhere on your path (Pipx is a good way to install python programs) and then run the command hatch build wheel to generate an installable python wheel.

Building the library containers is accomplished using the provided Makefile. In particular, for a specific system there will be a system container and possibly a specialized gazebo container. As an example, to run PX4 simulations you will need to build the px4-firmware and px4-gazebo make targets if you cannot download the images.

Running The Examples

Since this library is built using Hatch, we can utilize the environment management functionality to help simply running the examples. Each example has a hatch script that can be executed by running the command hatch run examples:<script>. For example, the px4 example is run using the command hatch run examples:px4.