Advanced Edge Deployment: Abstracting Cyber-Physical Models via FMU Mastery

Abstract

Deploying cyber-physical models at the edge or in the cloud as software components is the key step of modelbased- design. Depending on run-time environment, an extensive customization often needs to be made. To streamline and facilitate the deployment of models and simulators in production, a unified framework is developed. The implementation utilizes functional mockup units (FMUs) as the executable binary for the models and JavaFMI as the simulation engine. Each model deployment is encapsulated inside a microservice with all the software dependencies, with communication realized through RabbitMQ. A generalized approach to manage the model namespace has been implemented, ensuring that the FMU executor remains agnostic to changes in both model and application, as long as the AsyncAPI specification includes a mapping of the model's input-output space to the protocol’s topics. Two examples are presented to illustrate the convenience and effectiveness of the proposed framework: a winch controller at the edge for oil and gas wireline operation and a wireline logging unit simulator in the Azure DevOps pipeline for software-in-the-loop testing.