Adaptable Transition Systems

Abstract

Abstract. We present an essential model of adaptable transition systems inspired by white-box approaches to adaptation and based on foundational models of component based systems. The key feature of adaptable transition systems are control propositions, imposing a clear separation between ordinary, functional behaviours and adaptive ones. We instantiate our approach on interface automata yielding adaptable interface automata, but it may be instantiated on other foundational models of component-based systems as well. We discuss how control propositions can be exploited in the specification and analysis of adaptive systems, focusing on various notions proposed in the literature, like adaptability, control loops, and control synthesis.

Reconciling White-Box and Black-Box Perspectives on Behavioral Self-adaptation

“…Gem models describe the possible behaviours of a system as trajectories through its phase space 5 while the system is interacting with an environment, i.e., a system consists of a preordered time structure T , a phase space X and a set of trajectories in T → X. 6 T might be as simple as the natural numbers with the usual order, while in our example the state space might be the position of each robot. Since they are used during the analysis or the high-level design of a system, the initial Gem models of a system typically have very little information about its internal structure.…”

“…1) and allow one to specify or inspect (part of) the internal structure of a system in order to offer a clear separation of concerns to distinguish changes of behaviour that are part of the application or functional logic from those which realise the adaptation logic. This section summarises the CoDa approach to adaptation and one of its formalisations, namely Adaptable Transition Systems [6].…”

“…Indeed, AIAs come equipped with a suitable composition operator, denoted ⊗, which essentially corresponds to lifting the composition operator on the underlying Interface Automata structure. We refer the interested reader to [6]. The choice of symbol ⊗ is not an accident: interpreting composition in Gem models as composition of AIAs is indeed part of our approach to reconcile both views on adaptation, as we shall explain later in Section 4.…”

“…This clearly enables the use of standard model checking techniques to analyse properties of adaptive systems (see e.g. the references and the discussion in [6]). …”

“…The interested reader is referred to [7,10,23]. For a discussion of work related to the main formalisms used in this paper like Gem [14] and AIAs [6] we refer to the corresponding publications. We briefly discuss here some works related to properties of adaptive systems, a key notion in the presented contribution.…”