Ensemble-Based Modeling Abstractions for Modern Self-optimizing Systems

Abstract

In this paper, we extend our ensemble-based component model DEECo with the capability to use machine-learning and optimization heuristics in establishing and reconfiguration of autonomic component ensembles. We show how to capture these concepts on the model level and give an example of how such a model can be beneficially used for modeling access-control related problem in the Industry 4.0 settings. We argue that incorporating machine-learning and optimization heuristics is a key feature for modern smart systems which are to learn over the time and optimize their behavior at runtime to deal with uncertainty in their environment.

Comparing perfomance abstractions for collective adaptive systems

“…For instance, Lee et al [21] present a language, based on Architecture Analysis & Design Language (AADL), for drones that collaborate in packet delivery, specifically each drone needs to adapt to the motions of the other drones for collision avoidance. Töpfer et al [36] rely on modelling abstractions that incorporate machine-learning and optimisation heuristics to deal with uncertainty in the environment of CAS. The analysed scenario consists of workers going to a factory that may encounter delays and are replaced by standby workers.…”

Online ML Self-adaptation in Face of Traps

“…To experiment with SA systems and online ML, we have developed a machine-learning-enabled component model ML-DEECo [6]. Building on the DEECo ensemble-based component model [10], ML-DEECo enhances it with abstractions for estimating the future states (using supervised ML).…”

“…As a key entity supporting the self-adaptation of the system at hand, an ensemble is a dynamically formed group of the system's components serving for their cooperation. In this work, we focus only on the way online ML is employed in the adaptation rules in both components and ensembles (for more details on these two concepts, we refer the reader to our previous works [6], [10]). A key abstraction to support online ML in ML-DEECo adaptation rules is estimator.…”

“…What are the specific traps of applying online supervised ML for SA (experience using ML-DEECo)? To experiment with online supervised learning in selfadaptive systems, we developed a machine-learning-enabled component model called ML-DEECo [6] that allows modeling systems with ML-based predictions of the future. Alongside the component model, we provide its implementation as a Python framework [7] to facilitate prototyping and experimenting with the systems.…”

“…We have used the ML-DEECo framework to implement simulations of several self-adaptive systems including use-cases from smart farming [8] (one of them we use in this paper as the running example) and Industry 4.0 [6].…”