Design and analysis of time-critical parallel applications on heterogenous avionics platforms

Purpose and goal

The never-ending growth in the number of functions as well as their complexity in an aircraft, including autonomous behaviour based on machine learning (ML), will require an enormous amount of resources. At the same time, intelligent systems need to be functionally correct and trustworthy. The project aims at a systematic design methodology that can efficiently deal with the combination of the available parallelism in the applications, guarantees for non-functional requirements, and the emerging heterogeneous parallel platform architectures.

Expected effects and result

The project will develop a set of methods to exploit parallelism and at the same time guarantee timing properties: (a) a modelling approach that combines models of computation and data-parallel skeletons to yield a formal, executable, and analysable application model, (b) a timing analysis method to provide timing guarantees when executing on heterogeneous parallel platforms, and (c) a method to efficiently map application models to selected target platforms.

Planned approach and implementation

The project will study an avionics use case with elements of ML. The aircraft functions will be modelled, analysed and mapped to selected target platforms using the techniques developed in the project. The project will utilize results of previous Vinnova projects. KTH will research the underlying foundations of the methods in close contact with Saab, which defines future needs, provides industrial experience, and ensures the industrial relevance of the project.