Skip to content
/ csaf Public

Control Systems Analysis Framework - a framework to minimize the effort required to evaluate, implement, and verify controller design (classical and learning enabled) with respect to the system dynamics.

License

Notifications You must be signed in to change notification settings

GaloisInc/csaf

Repository files navigation

Control Systems Analysis Framework (CSAF)

CI

CSAF is a framework to minimize the effort required to evaluate, implement, and verify controller design (classical and learning enabled) with respect to the system dynamics. Its key features are:

  • Component based controller design
  • Native support for python and C language executables
  • Compatibility with external hardware and software processes
  • Ease of deployment

csaf_importing_components

Controllers, subsystems and plants are implemented as a collection of components. Below is an example of a topology graph of F16 system with GCAS autopilot.

f16_components

Quick Start

Installation

csaf_quickstart

Native

Clone the repo, and install via pip,

pip install csaf-controls

To install additional dependencies needed for the examples, go to the repo root and run

pip install -r requirements.txt

Docker

CSAF runs inside a Docker container, and in order to use CSAF, you first need to install docker. CSAF has been tested on Linux (Ubuntu 18.04 and 20.04) and OS X, but should run on any nix-like system that runs docker.

At the repo root, build the docker image

docker build -t csaf .

Now you can run an example

docker run -v $PWD:/out csaf python -m csaf_f16 --output-dir /out

Alternatively, a jupyter instance can be launched

docker run -p 8888:8888 csaf jupyter notebook --allow-root

Examples

CSAF currently contains a number of examples, including the F-16 shown below. These examples are packaged into component libraries. These libraries have application entry-points for demonstration. Note that CSAF has utilities to generate these entrypoints for user created component libraries.

F-16 Control System

To run a F16 system simulation, run

$ python -m csaf_f16 --time-max 30.0 --system F16Simple  

f16_components

The library CLI options are

$ python -m csaf_f16 --help

F16 Components (CSAF Library)
usage: __main__.py [-h] [-s SYSTEM] [-o OUTPUT_DIR] [-t TIME_MAX] [-d DO_OPEN]

F16 Components (CSAF Library) CSAF F16 Systems Viewer

optional arguments:
  -h, --help            show this help message and exit
  -s SYSTEM, --system SYSTEM
                        CSAF System to Simulate (examples are ['F16Simple', 'F16AirspeedSimple', 'F16Shield', 'F16MultiAgentCentral', 'F16AcasShield',
                        'F16AcasIntruderBalloon', 'F16AcasShieldIntruderBalloon', 'F16AcasShieldAcasIntruderBalloon'])
  -o OUTPUT_DIR, --output-dir OUTPUT_DIR
                        Directory to Store Artifacts
  -t TIME_MAX, --time-max TIME_MAX
                        Timespan to Simulate
  -d DO_OPEN, --do-open DO_OPEN
                        Open Simulation Result

Jupyter notebooks

CSAF can be used from within a jupyter notebook. To start CSAF in the notebook mode, go to the notebooks folder and run

jupyter notebook

CLI Apps

CSAF has functionality to wrap CLI functionality around core CSAF concepts, like systems and scenarios. For example, after creating a scenario type AcasSimpleScenario, one can create an app by adding

from csaf_f16.goals import AcasSimpleScenario
from csaf.utils.app import ScenarioCsafApp


if __name__ == '__main__':
    app = ScenarioCsafApp(AcasSimpleScenario)
    app.main()

Passing in the scenario.json file

{
  "rel_pos_x" : 5000,
  "rel_pos_y" : 5000,
  "rel_angle" : 0.5,
  "rel_speed" : 200
}

with

python scenario_app.py -i scenario.json -t 200.0

creates traces.json file with simulation results. Also, docker can be used, such as,

docker run -v $PWD:/out csaf python /out/scenario_app.py -i /out/scenario_list.json -o /out/docker_traces.json

Tests

CSAF unit and integration tests use the pytest framework. After installing the mypy plugin, static tests can be added by adding the switch

pytest --mypy -s tests

Development

CONTRIBUTING.md contains CSAF development guidelines, please familiarize yourself with the guidelines before opening a pull request. The best way to contact the dev team is via GitHub issues.

Licensing

The code in this repository is licensed under two different licenses. The core of CSAF (csaf directory) and the majority of examples is licensed under BSD license. The f16 examples in the csaf_f16 module is licensed under GPL license.

Acknowledgment

This material is based upon work supported by the DARPA Assured Autonomy program under the United States Air Force under Contract No. FA8750-19-C-0092. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of DARPA or the United States Air Force.

Distribution

"Distribution Statement "A" (Approved for Public Release, Distribution Unlimited)” "© 2021 Galois, Inc."

About

Control Systems Analysis Framework - a framework to minimize the effort required to evaluate, implement, and verify controller design (classical and learning enabled) with respect to the system dynamics.

Topics

Resources

License

Stars

Watchers

Forks

Languages