Section: New Results

Cross layer optimization

The desired behavior of embedded systems is usually subject to real-time constraints and it involves interactions with a set of physical devices or components, each with its own individual constraints. The general formalism of timed automata allows expressing the expected software behavior of a device in the form of a finite state automaton extended with real-valued variables called clocks. The particular constraints of physical devices themselves can also be expressed in this form. In this context, our goal is to provide an optimized mapping of a software protocol on a hardware device by taking advantage of all various low-power consumption states when possible, thus taking a cross-layer approach on the problem of energy consumption minimization. Generally the software to hardware mapping does not take benefit from all low-level functionality since it implies to ensure complex time constraints. In [14] we address the problem of mapping the free states of a software protocol expressed as a timed automaton to a physical device whose behavior is also expressed as an automaton with states of fixed or unbounded (but with lower limit constraint) duration. We propose a methodology that allows a mapping of those free states to unique states or paths between states in the device automaton with distinct edges, that guarantee that all the time constraints are met and all feasible transitions in the protocol automaton remain realizable, all while minimizing the energy consumed.