Learning-enabled components are increasingly used in CPS applications but present unique challenges for development, testing, and verification. Existing tools do not provide integrated support for the entire development lifecycle of CPS with LECs, from initial design through safety assurance. This toolchain addresses that gap by enabling structured, reproducible development processes and comprehensive documentation of ML component provenance.
This demonstration paper presents an integrated toolchain for developing Cyber-Physical Systems with Learning-Enabled Components (LECs). The work showcases a comprehensive workflow built on the WebGME platform that supports architectural modeling, component library management, automated data collection, and performance evaluation for machine learning-based CPS components.
The paper demonstrates the ALC (Assurance-based Learning-enabled CPS) toolchain through an autonomous underwater vehicle control example. The system supports design of multiple implementation alternatives, comparison of CNN architectures for control tasks, and integration of both supervised and reinforcement learning approaches within a unified model-based development environment.
@inproceedings{Hartsell2019a,
author = {Hartsell, Charles and Mahadevan, Nagabhushan and Ramakrishna, Shreyas and Dubey, Abhishek and Bapty, Theodore and Karsai, Gabor},
booktitle = {Proceedings of the 10th {ACM/IEEE} International Conference on Cyber-Physical Systems, {ICCPS} 2019, Montreal, QC, Canada},
title = {A {CPS} toolchain for learning-based systems: demo abstract},
year = {2019},
pages = {342--343},
abstract = {Cyber-Physical Systems (CPS) are expected to perform tasks with ever-increasing levels of autonomy, often in highly uncertain environments. Traditional design techniques based on domain knowledge and analytical models are often unable to cope with epistemic uncertainties present in these systems. This challenge, combined with recent advances in machine learning, has led to the emergence of Learning-Enabled Components (LECs) in CPS. However, very little tool support is available for design automation of these systems. In this demonstration, we introduce an integrated toolchain for the development of CPS with LECs with support for architectural modeling, data collection, system software deployment, and LEC training, evaluation, and verification. Additionally, the toolchain supports the modeling and analysis of safety cases - a critical part of the engineering process for mission and safety critical systems.},
bibsource = {dblp computer science bibliography, https://dblp.org},
biburl = {https://dblp.org/rec/bib/conf/iccps/HartsellMRDBK19},
category = {poster},
contribution = {colab},
doi = {10.1145/3302509.3313332},
file = {:Hartsell2019a-A_CPS_Toolchain_for_Learning_Based_Systems_Demo_Abstract.pdf:PDF},
keywords = {cyber-physical systems, machine learning, model-based design, toolchain, learning-enabled components},
project = {cps-autonomy},
tag = {ai4cps},
timestamp = {Sun, 07 Apr 2019 16:25:36 +0200},
url = {https://doi.org/10.1145/3302509.3313332}
}