Dependable platforms for autonomous systems and control
DPAC aims to establish a leading research profile targeting dependable platforms for autonomous systems and control. This will be accomplished through close collaboration and interaction between Embedded Systems (ES) research groups at MDH and the participating industrial companies. DPAC is hosted at Mälardalen University in the ES research environment.
Embedded computer systems are nowadays incorporated in many kinds of products including many mission critical applications such as trains, autonomous utility vehicles, aviation, smart grid power management etc. These systems offer advanced functionality and fulfil an important role for the competitiveness of companies and the future national and global infrastructure. The scientific and technical results of DPAC will support future innovation by providing dependable platforms that can be used to efficiently realise dependable, reliable and safe electronically controlled products.
Four established research groups from MDH will, in addition to the staff from companies, provide the core competence thrust within DPAC. The research will be organised around three main research areas:
- Predictability and dependability in parallel architectures
- Autonomous systems and control
- Design methodologies
These combined competences give DPAC a unique opportunity to address system-wide research challenges that span several traditional research areas and wide industrial applications as well as forming a robust basis for the research in DPAC.
Projects within DPAC
When launched in October 2015, DPAC started with three larger projects, one for each research area that will tackle a concrete subset of the overarching research agenda. These projects each serve the dual purpose of providing tangible research results together with our industrial partners, and to build a strong reputation for MDH in the area of dependable platforms. Furthermore, a DPAC unifying use-case ties the projects together.
Predictability and dependability in parallel architectures
In this project we will develop improved software support for reconfigurable and dependable use of parallel architectures, ranging from contemporary multiparallel multicores to future hyperparallel heterogeneous platforms such as specified by the standard Heterogeneous Systems Architecture (HSA). The major research challenge is the fact that real-time behaviour has not been a driver of recent hardware development; only performance has been. Hence, in order to make use of parallel architectures it is of essence to understand and define execution models for real-time dependable software based upon these architectures.
Therefore we will develop operating-system support for allocating computing resource to dependable and timing-sensitive computations and support for policing execution according to these allocations. We need to manage allocation and policing of all hardware resource that are needed for computations, including CPU bandwidth, memory bandwidth, cache areas, GPU clusters, DPS clusters, and FPGAs.
Autonomous systems and control
The overall objective of this project will be to develop autonomous platform solutions that have the ability to judge and decide its own level of autonomy and still be able to meet the dependability and functionality requirements of the system. To demonstrate and verify this novel approach a dependable autonomous platform for safe and secure path following in a real-world setting will be developed. A mission management tool will be designed that allows seamless interaction between autonomous systems as well as theoretical work consisting of a classification scheme for defining different levels of autonomy which also incorporates dependability.
Highly integrated dependable systems call for structural design methods with key driving forces representing dependability requirements and constraints. In this project we establish a design framework for creating and analysing dependable platform designs by providing methodologies capturing dependability requirements of modern autonomous and control systems.
Based on a unifying use case and other potential use cases from partner companies, methods will be developed and evaluated. We do this in three work packages:
- Requirements capture and validation,
- Design assurance and
- Design analysis