Implementing Timed Automata Specifications: The "Sandwich" Approach

Abstract

From a highly distributed timed automata specification, the paper analyses an implementation in the form of a looping controller, launching possibly many tasks in each cycle. Qualitative and quantitative constraints are distinguished on the specification to allow such an implementation, and the analysis of the semantic differences between the specification and the implementation leads to define an overapproximating model. The implementation is then "sandwiched" between the original specification and the new model, allowing to check if the important properties of the specification are preserved by the implementation.

Indefinite waitings in MIRELA systems

“…Rendering R when a synchronisation is possible, time may not progress. More precisely, we used a subclass called Timed Automata with Synchronised Tasks (TASTs) in order to cope with implementability issues (see [7] for more details). Syntactically, a TAST is an annotated directed (and connected) graph, with an initial node, provided with a finite set of non-negative real variables called clocks (e.g., x), initially set to 0, increasing with time and reset (x := 0) when needed.…”

“…The aim of this paper is to provide a formal method support for the development of concurrent applications, which consist of components, which mutually interact in a way, that should meet certain real-time constraints, like a reaction time within a given time period. MIRELA (for MIxed REality LAnguage [9,7,10,8]) was initially meant to be used for developing mixed reality (MR) [6] applications, which acquire data from sensors (like cameras, microphones, GPS, haptic arms. .…”

“…The second phase concerns the analysis of the system: it essentially consists in analysing through model-checking a set of desired properties considered important, either the absence of bad behaviours or the satisfaction of timing constraints. In the third phase, such a checked specification is used to produce an implementation skeleton, in the form of a looping controller parametrised with a sampling period and possibly executing several actions in the same period, aiming at preserving those properties [7]. We revisit here essentially the second phase of the methodology of the MIRELA framework.…”

Deadlock and Temporal Properties Analysis in Mixed Reality Applications

“…The MIRELA framework [7] proposes a methodology that consists of three phases: In the first phase, a formal specification of the system in the form of a network of timed automata [1] is built. It may be obtained by a translation from a high level description made of connected components, and represents an ideal world, for which some properties considered important may be checked.…”

“…Since a few years, the MIRELA framework (for MIxed REality LAnguage [8], [7], [16], [9]) is developed aiming at supporting the development process of applications made of components which have to react within a fixed delay when some events occur inside or outside the considered area. This is the case in mixed reality applications which are evolving in an environment full of devices that compute and communicate with their surrounding context [6].…”