Section: New Results

Design-driven Testing by simulation

Previously, we have introduced a paradigm-oriented development approach that revolves around a conceptual framework concretized by a design language  [26] . A design description is used to generate high-level programming support, to perform a range of verifications, and to abstract over underlying technologies.

This approach is illustrated with the Sense-Compute-Control (SCC) paradigm  [48] , where an SCC software system gathers information about an environment via sensors (whether hardware or software) and issues orders to impact the environment via actuators. The SCC paradigm has a wide spectrum of applicability; we have used it successfully in the domains of home/building automation, multimedia, avionics and networking.

SCC systems involve both software concerns, like any software system, and integration concerns, for the constituent networked entities forming the environment of the SCC-loop. This situation is problematic for testing because it requires acquiring, testing and interfacing a variety of software and hardware entities. This process can rapidly become costly and time-consuming when the target environment involves many entities.

We have developed a simulation approach and a tool named DiaSim that leverage the DiaSpec description of an environment [15] . This description is used to generate both a programming framework to develop the simulation logic and an emulation layer to execute applications. The generic nature of our approach has been illustrated by leveraging two different simulation tools, namely, Siafu for 2-D rendering of home/building spaces, and FlightGear for avionics.

To fuel the simulation of an environment with accurate stimuli, we need to model real systems, including natural phenomena (e.g., heat transfer in a building) or mechanical systems (e.g., aircraft models). These physical models are typically defined as continuous systems using differential equations. To facilitate the reuse of off-the-shelf physical models, we have used a DSL named Acumen for describing differential equations. Acumen continuous models are coupled with the DiaSim discrete simulator, forming a hybrid system fueled by accurate stimulus producers [18] .

These major accomplishments were conducted by Julien Bruneau, in the context of his PhD studies [11] .