This is an artifact repository for ACSOS 2022 paper ML-DEECo: a Machine-Learning-Enabled Framework for Self-organizing Components by Michal Töpfer, Milad Abdullah, Martin Kruliš, Tomáš Bureš, and Petr Hnětynka.
This repository features implementation of the ML-DEECo framework as well as two examples of simulations implemented using the framework.
ml_deeco– a framework for implementation of simulations of adaptive distributed systems with native support for machine learningdrone_charging_example– simulation of a smart farming system for protection of fields against birds using battery-powered flying dronessmart_factory– simulation of a smart factory with adaptive security rules and machine-learning-based estimates of workers' tardiness
The ML-DEECo itself requires Python 3.6+ and the following packages installable by
pip install numpy seaborn matplotlib pyyaml
The typical expected application is to include the ML-DEECo repository as a submodule of your simulation repository (as we did in this artifact).
The artifacts may require additional packages based on which ML library they are using (detailed in their readme files).
The documentation is scattered in several readme files:
- ML-DEECo readme – describes the installation and usage of the framework itself. Also explains the components and ensembles as well as overall architecture.
- ML-DEECo simple-example readme – explains the simple example which can be used as a bootstrap code to quickly sketch your own simulation
- Drone charging example readme – contains the installation and usage guidelines for the smart farming example, expected results, configuration details, and component model overview
- Factory workers example readme – contains the installation and usage guidelines for the factory access control example, expected results, and simulation architecture overview