The focus of the AstréeA project is on the development of static analysis by abstract interpretation to check the safety of large scale asynchronous embedded software. The safety of critical embedded software, such as found in aeronautical, railway, automotive or medical industries, is of paramount importance as any error may have catastrophic economical or human consequences. The cost of testing, the main validation method in use, is unfortunately becoming prohibitive when faced with the continually increasing complexity in embedded systems. Thus, tests only cover a small fraction of all possible executions, and can fail to discover some errors during validation. The situation is critical in the case of asynchronous software, i.e., software composed of several processes executed in parallel, as tests can only cover a tiny part of the huge number of process interleavings, while some errors may appear only for some specific, hard to reproduce interleaving. Formal methods, however, enable validation with mathematical guarantees. Static analysis by abstract interpretation seems ideally suited to an industrial context as it is automatic (without user intervention), works directly on the source code (and not a simplified model), is sound (full coverage of the formal semantics of the language, hence without false negative) and efficient. However, it can suffer from false alarms (false positives), which is the case of most commercial static analyzers. Our experience with Astrée (http://www.astree.ens.fr/) has shown that it is possible to build a static analyser with zero false alarm (giving thus a proof of safety) if it is specialized for a class of properties and programs: in the case of Astrée, run-time errors in synchronous embedded control/command avionic software. Astrée is now a commercial product used in an industrial context. In the AstréeA project, we focus on the, much harder, issue of run-time errors in asynchronous software, by developing the AstréeA prototype analyzer, based on Astrée. Following the development method used successfully for Astrée, AstréeA is also specialized, in this case for run-time errors in large scale asynchronous avionic software running under a real-time operating system. Airbus France provides a large typical example of such software (1.6 M lines). During the first phase of AstréeA, funded by ANR, we developed a concrete and abstract models of the ARINC 653 operating system and its scheduler, and a first analyzer prototype. After 4 years, it was able to analyze the full software, albeit with many alarms. The gist of the AstréeA proposition is the continuation of this effort, following the recipe that made the success of Astrée: an incremental refinement of the analyzer until reaching the zero false alarm goal. For AstréeA, the refinement concerns: (1) the abstraction of process interactions (relational and history-sensitive abstractions), (2) the scheduler model (supporting more synchronisation primitives and taking priorities into account), (3) the memory model (supporting volatile variables), and (4) the abstraction of dynamical data-structures (linked lists). These improvements should enable AstréeA to achieve a high level of precision, and even the zero false alarm goal, thus making its industrial use possible.

In the field of transportation, vehicles (ground or air) are becoming more and more autonomous and communicating. The emergence of these new technologies may relegate pilots to a supervisory role, implying less vigilance and less awareness of the environmental context. This degraded state can hinder the effective recovery of the vehicle and lead to dangerous situations. The objective of the COMMUTE project is to develop a genuine non-verbal, multimodal, intuitive and interactive communication system between the driver and his vehicle. For this purpose, multimodal solutions based on a cognitively situated approach will be developed within the framework of an interactive multimodal synthesis platform. Two use cases, presenting a strong safety issue, will constitute the common thread on which theoretical and experimental developments will be based: emergency warning (short time) and continuous regulation (long time).

NextSim partners, as fundamental European players in Aeronautics and Simulation, recognize that there is a need to increase the capabilities of current Computational Fluid Dynamics tools for aeronautical design by re-engineering them for extreme-scale parallel computing platforms. The backbone of NextSim is centered on the fact that, today, the capabilities of leading-edge emerging HPC architectures are not fully exploited by industrial simulation tools. Current state-of-the-art industrial solvers do not take sufficient advantage of the immense capabilities of new hardware architectures, such as streaming processors or many-core platforms. A combined research effort focusing on algorithms and HPC is the only way to make possible to develop and advance simulation tools to meet the needs of the European aeronautical industry. NextSim will focus on the development of the numerical flow solver CODA (Finite Volume and high-order discontinuous Galerkin schemes), that will be the new reference solver for aerodynamic applications inside AIRBUS group, having a significant impact in the aeronautical market. To demonstrate NextSim market impact, AIRBUS has defined a series of market relevant problems. The numerical simulation of those problems is still a challenge for the aeronautical industry and their solution, at a required accuracy and an affordable computational cost, is still not possible with the current industrial solvers. Following this idea, three additional working areas are proposed in NextSim: algorithms for numerical efficiency, algorithms for data management and the efficiency implementation of those algorithms in the most advanced HPC platforms. Finally, NextSim will provide access to project results through the “mini-apps” concept, small pieces of software, seeking synergies with open source components, which demonstrate the use of the novel mathematical methods and algorithms developed in CODA but that will be freely distributed to the scientific community.