Real-Time Adaptive Abstraction and Approximation Using Validity Frames -- an Experience Report

Abstract

Designing a Cyber-Physical System (CPS), including modeling the control components and services, is a challenging task. Using models and simulations during run-time is crucial for successfully implementing advanced control and prediction components. The complexity of designing an effective CPS system increases due to real-time constraints. Generating accurate predictions and making decisions using detailed models in various contexts is %can be computationally demanding and complex to manage within the available computational resources. Employing approximated models and switching to the most suited model adaptively at run-time is an effective technique. But an approximated model is most probable not valid in all the different contexts the system will be in. This experience report uses the Validity Frame concept to enable %this adaptation at run-time. In each environment, some influencing factors are outside the model's control, but these properties influence the model's behavior. By defining Validity Frames, based on specific contexts and related models, we present a possible perspective to address the issue of selecting the more appropriate model in various contexts. Furthermore, we discuss the insights and lessons obtained and determine future challenges.