The Assurance-based Learning-enabled Cyber-Physical Systems (ALC) toolchain is an integrated set of tools and corresponding workflows specifically tailored for the model-based development of Cyber-Physical Systems (CPS) that utilize Learning-Enabled Components (or LECs). 

ALC toolchain supports:

• Modeling:  A WebGME based modeling environment that supports SysML-style models and their extensions that include architecture models, requirements model, functional decomposition models, dynamic risk models and hazard models as well as cross-linking these models to capture their inter-dependencies. 

  Image

  

• Execution: extensive support for model construction, engineering, and integration of LECs and assurance technologies, including: 

  • “Headless” execution of simulations for training data collection, LEC training, evaluation, and verification.

  • Setup and execution of campaigns over parameter spaces of interest.

  • Orchestration of workflows related to development, testing, and evaluation of LECs and related assurance technologies.

  • IDE/VNC/ git/Docker:  web based Integrated Development Environment (IDE) for interactive code development; debugging and testing of the software components including support for viewing graphical user interfaces using web based VNC; git server to host software repositories for the version control of software components; docker registry to store and share the images between the interactive environment and the headless execution.

  • Support for extension points to integrate new simulation environments, associated software tools for scenario/ data generation, deployment, and testing.

  Image

  

• Assurance technologies: 

  • Dynamic run-time assurance monitors that can detect, at run-time, Out-of-Distribution anomalies of the inputs (and/or output/s) of the LECs.

  • ReSonAte based computation of metrics related to static and dynamic assurance metrics for hazard mitigation in the threat propagation paths captured in the dynamic risk evaluation models.

  • Real-time reachability analysis to predict safe and unsafe operation of the LECs  in operating environments.

  • Behavioral-tree based support for contingency management for mission execution.

• Assurance cases: 

  • modeling and analysis of safety cases with Goal Structuring Notations (GSN) and their cross-referencing to SysML models and evidences for explanation and traceability.

Image

The toolchain includes a BlueROV2 based UUV demonstration platform to highlight the above technologies.