KTH Fernando Herrera Site

Topics
My background at that time was mostly on system-level modelling and specification methodologies for ESL design, and the development of ESL design methods (e.g. design space exploration, embedded software generation, etc). In KTH/ICT/ES my research interest was more oriented to a coherent integration of system-level specification methodologies, ESL specification design techniques and model-driven development concepts and techniques. The sinergy of these topics is necessary for building advanced design methodologies which are being demanded, specifically in the domain of Mixed Criticality Systems (MCS) and Systems of Systems (MCSoS).
Read more

Mixed Criticality systems (MCS) are systems which have design constraints at very different levels of demmand. For instance, a mobile phone has tasks which are definitely more important than others, e.g. an incomming call should take over the screen and the loudspeakers if we are playing bubbles. MCS usually mix real-time requirements with functional equirements. MCS have been pushed by a growing trend to build systems through the integration of systems, which is called SystemS of SystemS (SoS). For instance, a car or plane integrate different types of systems and networks, e.g., driving assistance, multimedia, navigation systems in a car, or the flight control system and the multimedia system for passangers in a plane. Traditionally, this systems have separated resources and architectures. However, it tremendously raises the costs. Solutions pushing more sharing of resources will bring more cost-effective solutions, which, by scale economy, not only will pusht cheaper solutions, but it can also leverage a democratization of safety electronics. However, to reach that point, it is still necessary to define design methodologies capable to integrate functionalities of many different criticalities and functional and performance demands in the most cost-effecient platform, without sacrifying safety and average performance. It will likely require to rely on heterogeneous platforms, integrating predictable resources and resources optimized for optimum average performance. The vision is a tooling that can let the system enginner to decide the architecture of, e.g. a set of embedded devices interconnected through a TTEthernet and maybe other more cost-efective buses in a car running control and entertainment applications. MCSoS design will be about optimizing, for instance, the resources required for the video reproduction and distribution to backseats, while at the same time the deadlines on safety critical control applications are guaranteed taking into account that many of those devices and resources, such as the communicating infrastructure (in this case, the TTEthernet bus) is shared.

MCS and MCSoS design requires to adquire a global view and merging of many disciplines: Model Driven Development (MDD), System-Level modelling, Models of Computation, Electronic System-Level (ESL) design, Schedulability analysis, Network modelling and analysis, etc. It is challenging, but appealing too!.

Mission
My first task in KTH/ICT/ES was the definition of basic lines for MCS and MCSoS modelling and design. I also focused on the integration of technologies already developed by KTH/ES, and open methods&tools available to define MCS/MCSoS methodologies. In those activities I also had in mind the definition of a research track whch pushed the collaboration of KTH/ES with my original research group, the Microelectronics Engineering Group of the University of Cantabria, where I graduated as Ph.Dr in 2009, and where I developed my former 11 years of research work (see my site in UC/GIM), and with other partners, like PoliMi, i had collaborated in the past with too. Read more

KTH/ES and UC/GIM expertises had quite complementary know-how. In UC, we were working in developping Model Driven Methodologies relying on UML/MARTE modelling methodologies and we used the ForSyDe methodology to provide a formal background to them. Strengthening this line was yet an objective of my research activity here. However, after updating my knowledge of the current research status of KTH/ES, we found that UC/GIM and KTH/ES have many other different and complementary backgrounds and expertises in ESL which fit well and could be exploited through immediate and future collaboration in the definition of modelling and design techniques for MCS and MCSoS systems.

After having defined basic research lines for MCS and MCSoS modelling and design (reported in DSD'13), the main focus of my research in the KTH/ES group ihas been put on one of those lines, the development of a novel Joint-Analytical and Simulation-based framework for Design Space Exploration (JAS-DSE) . The feasibility and main benefits of the framework was shown in a paper in FDL 2013. In FDL13, The static analysis part was demonstrated by relying on a simple MiniZinc coding of the DSE problem. The simulation-based part relied on the KisTA framework.

FDL'13 provided the demonstration of the concepts. Then we focused on the development of a framework which could tackle more generic and realistic examples, and which facilitiates the live of the user, by giving a user interface customized to a system-level designer. The development of such complex JAS-DSE framework required the integration of several KTH/ES tools&frameworks.The static part of the JAS-DSE flow relies on an Analytical-DSE tool developed by my colleage Kathrin Rosvall. This tool is based again on constrain programming (but relies on Gecode, instead of MiniZinc) and is capable to capture more generic models than in FDL13 work and to abstract the modelling part, considering communication times.

That enables to start a collaboration with PoliMi (since around October 2013 until April 2014) which resulted in a first version of the framework which confirmed the advances of the benefits advanced in FDL'13,
and which provides the aformentioned major benefits.

KTH and the Deib lab. of the Politecnico di Milano, PoliMi/Deib, we have performed a next step by integratin the MOST tool in the JAS-DSE framework. The work done with Edoardo Paone and Gianluca Palermo has been an important step, because it meant the solution of novel technical issues (which can appear in general in any two-phased JAS-DSE framework as the one proposed), and at the same time it enables important benefits for making the JAS-DSE framework applicable to complex embedded systems.

In that work we have to solve certain challening issues, which i do not mention yet here since this work is pending for publication.

Current work of his JAS-DSE framework is its connection to system-level models, specifically ForSyDe-SystemC models, enabled by my colleaghe Hosein Attarzadeh.

The KisTA framework oriented to the time analysis of predictable systems. In this framework I combined advanced features I knew from the work of my colleagues in UC, plus some features which were interesting for our work in KTH, and which results in a novel and powerful combination of features, Such features have been reported in FDL'04 (to appear),

At the beginning of my research here I was assesseing the posibility to integrate ForSyDe in a model-driven development environment. While this work was progressing and it is reported, the biggest priority of the aforementioned activities prevented me from further progress.

In the last period of my research stay in KTH, I worked in cooperation with the Microelectronics Engineering Group of the University of Cantabria again, in the definition of a MARTE metamodel, which is crucial for enabling efficient design of MCS/MCSoS.

- F. Herrera and I. Sander. "An extensible Infrastructure for Modeling and Time Analysis of Predictable Embedded SystemS". To Appear in FDL'14.

I will also report in this site at the publication related to the development of the first version of JAS-DSE framework done in KTH, and which is still work in progress.

- F. Herrera (KTH), J. Favaro (INTECS), C. Lomba (GMV). "Safety Standard Assessment" v0.5. Internal document of the CONTREX projet (T2.1), May 13th, 2014.

- F. Herrera, H. Attarzadeh, I. Sander (KTH) and J. Medina (UC). "Formal UML/MARTE modelling relying on Models of Computation&Communication (MoCC): An assessment considering ForSyDe and HetSC methodologies". v0.4. Jun. 25th. 2014. Internal document of the CONTREX project. Task 2.1.

- J.Medina (et.al) "Design of embedded mixed-criticality CONTRol systems under consideration of EXtra-functional properties". CONTREX/UC/R/2.1.1 deliverable of the CONTREX prokect. June, 2014.

Here, I make available some modifications and pre-compiled tools which i have used for my research and which should make easier to anyone to reproduce the experiments published.

A modified version of the SCoPE v1.1.5 sources, which can be compiled in a 64 bits machine (checked on Ubuntu LT12.04) and against SystemC 2.3.

Stand-alone bynary versions of scope-g++ and opcost, which should work on any Linux-64 bits configuration.

KTH Personal site: Fernando, Herrera Casanueva (last update: 2014/07/03)