Researcher in charge: Dávid Juhász

Industrial supervisor: Stefan Blixt

Industrial host: Imsys AB, Sweden

Academic supervisor: Prof. Dr. Axel Jantsch

Academic host: Technische Universität Wien, Austria

Topic Description:

Cyber-Physical Systems are distributed computational systems which interact with their physical environment in a reactive way including some level of self-awareness.

In many scenarios, self-awareness is founded on the observations of the system, as in Observe-Decide-Act loops. Various observations, structured according to different aspects, build up an understanding of the status and behaviour of the system and its environment. Decision making processes utilize structured and annotated observational data for the system to act in a way that steers the system and its environment towards a desired state.

State-of-the-art systems are based on nodes manufactured with multi-core heterogeneous architectures. Heterogeneity of modern architectures covers a vast spectrum of different hardware configurations, among which are differently sized general-purpose cores with the same instruction set architecture, general-purpose cores with different instruction set architectures, and various accelerators providing efficient realization of special features. A heterogeneous architecture can be implemented as a shared memory system, as a system of cores with separate memory address spaces, or as a mixture of those. Though modern hardware platforms provide ever-growing computational power, they are and foreseeably remain resource-constrained systems. Meeting their tight energy budget whilst maintaining functional and non-functional requirements is of utmost importance.

On the one hand, the diverse computational capabilities of hardware platforms make it possible to realize complex and computationally demanding features in software. On the other hand, a number of logically separate software modules of a complex system are to be executed on one node, which is a heterogeneous hardware platform, so that available hardware resources would be exploited effectively. Those separate modules tend to be of different criticality levels, thus mixed-critical applications are executed on the majority of nodes in Cyber-Physical Systems.

ESR investigates the software issues of modern Cyber-Physical Systems with respect to the efficient utilization, with special interest in energy efficiency, of heterogeneous architectures by mixed-critical, reactive applications which are defined in the terms of self-awareness. The research is focused on the intra-node aspects of software execution, inter-node communication being a higher level concern of distributed systems is not taken into account. The utilization of dedicated execution nodes is considered. That is, the entire software application is statically known for code generation and no interference caused by executing statically unknown pieces of software is studied.

ESR reviews the state-of-the-art of related literature, then designs, implements, and evaluates a solution, which may consist of a compilation toolchain and runtime system, for implementing and reliably executing mixed-critical, self-aware cyber-physical applications on heterogeneous hardware platforms with respect to minimizing the overall power consumption of the execution node.

The research topic is tightly connected to the research lines RL2 and RL4. The planned software solution aims at coping with resource constraints posed by embedded platforms (RL2) and optimizing efficacy and reliability of executing mixed-critical software (RL4).

Presentations:

  • Dávid Juhász, Axel Jantsch, Maximilian Götzinger, Nima Taherinejad: Modeling Self-Awareness. Poster presentation at SelPhyS 2017, Self-Awareness in Cyber-Physical Systems. April 21, 2017; Lawrence Convention Center, Pittsburgh, USA. [poster, slideshow]

Secondments:

  • Technische Universität Wien, Austria (continous, 3 months/year, 2016-2019)

ESR participates in the work of the System-on-Chip research group of TUW ICT. He contributes to the development of the RoSA (Research on Self-Awareness) framework.

  • INCHRON GmbH, Germany (July 17 – August 11, 2017)

ESR studied the INCHRON Tool-Suite and investigated its applicability for his research. The simulation module of the INCHRON Tool-Suite, ChronSIM, is used by ESR to prototype and evaluate his scheduler.