ARTEMIS-2011-1Full Project ProposalDECISIVE
ARTEMIS Call 2011
ARTEMIS-2011-1
DECISIon and platform support for model‐based eVolutionary development of Embedded systems
Date of preparation:August 19, 2011
Version number(optional): 0.8
ARTEMIS Sub-programme addressed (see Annual Work Programme 2011 section 3.2)
Major: ASP1: Methods and processes for safety‐relevant embedded systems
Minor: ASP5: Computing platforms for embedded systems
Industrial Priority addressed see Annual Work Programme 2011 section 3.1)
Major: Design methods and tools
Minor: Reference designs and architectures
Name of the coordinating person: Frank van der Linden
e-mail:
List of participants:
Participant no. (1) / Participant organisation name / Part. short name / Country / ARTEMIS Member State(Y/N) / Other EU Member or State/Ass. country
(Y/N) / National eligibility checked by applicant (Y/N) (2)
1 (Coordinator) / Philips Medical Systems Nederland BV / PHILIPS / NL / Y / N / Y
2 / AVL List GmbH / AVL / AT / Y / N / Y
3 / CISC Semiconductor Design+Consulting GmbH / CISC / AT / Y / N / Y
4 / NXP Semiconductors Austria GmbH / NXP-A / AT / Y / N / Y
5 / Technical University of Denmark / DTU / DK / Y / N / Y
6 / PAJ Systemteknik / PAJ / DK / Y / N / Y
7 / Contribyte Oy / COY / FI / Y / N / Y
8 / Konecranes Heavy Lifting Corporation / Knr / FI / Y / N / Y
9 / Mega Electronics Ltd / MEGA / FI / Y / N / Y
10 / Nokia Siemens Networks / NSN / FI / Y / N / Y
11 / Convergens Oy / CON / FI / Y / N / Y
12 / University of Eastern Finland / UEF / FI / Y / N / Y
13 / University of Oulu / UoO / FI / Y / N / Y
14 / Technical Research Center of Finland / VTT / FI / Y / N / Y
15 / Atego SAS / Atego / FR / Y / N / Y
16 / Commissariat à l’EnergieAtomique et aux Energies Alternatives / CEA / FR / Y / N / Y
17 / European Aeronautic Defence and Space Company EADS France SAS / EADS / FR / Y / N / Y
18 / Valeo / Valeo / FR / Y / N / Y
19
20 / Christian-Albrechts-Universitätzu Kiel / CAU / DE / Y / N / Y
21 / FraunhoferGesellschaftzurFörderung der angewandtenForschunge.V. / FHG / DE / Y / N / Y
22 / NXP Semiconductors Germany GmbH / NXP-D / DE / Y / N / Y
23 / Centro Ricerche Fiat S.C.p.A / CRF / IT / Y / N / Y
24 / Computers Guard / CG / LV / Y / N / Y
25 / Latvian Railway / LDZ / LV / Y / N / Y
26 / Riga Technical University / RTU / LV / Y / N / Y
27 / Almende / ALM / NL / Y / N / Y
28 / Océ Technologies BV / OCE / NL / Y / N / Y
29 / TechnischeUniversiteit Eindhoven / TUE / NL / Y / N / Y
30
31 / Ikerlan-IK4 / IKER / ES / Y / N / Y
32 / Integrasys / ISYS / ES / Y / N / Y
33 / MondragonUnibertsitatea / MU / ES / Y / N / Y
34 / ULMA Embedded Solutions / UES / ES / Y / N / Y
35 / Mälardalen University / MDH / SE / Y / N / Y
36 / Volvo / Volvo / SE / Y / N / Y
37 / Arcticus Systems AB / ARC / SE / Y / N / Y
38 / Hoxville Oy / HOX / FI / Y / N / Y
39 / TechnischeUniversiteit Delft / TUD / NL / Y / N / Y
40
41
42 / ETAS GmbH / ETAS / DE / Y / N / Y
43 / Fondazione Bruno Kessler / FBK / IT / Y / N / Y
(1) Please use the same participant numbering as that used in Proposal submission forms A2
(2) For partners from ARTEMIS Member States, please indicate whether you consider that you comply with the national eligibility criteria for funding as stated in the document "Eligibility Criteria" published in the Call.
Proposal abstract
(copied from Part A)
The objective of the project is to develop a methodology and tool support for model-driven evolutionary development of complex embedded systems. Supporting evolutionary design will reduce the development time and time-to-market, reduce development and unit costs, increase the quality of the products and of the engineering processes and reduce re-certification costs.
Today, most systems are engineered in an evolutionary fashion: introducing a new version of an existing product, introducing new features—possibly as part of an evolution of a product line, performing a design-iteration, etc. The models of an embedded system will evolve at the same time with the system. However, none of the current state-of-the-art approaches to model-based engineering of embedded systems support evolutionary development.
The project deploys the model-based evolutionary development in domains with the following characteristics:
•Longevity, addressing the concerns of high quality, product evolution, platform and low-cost maintenance
•Tuning & Scaling, addressing the concerns of physical variations in production process, high configurability and need for calibration
•Reliability Safety, addressing the concerns of measurement & control, safety critical systems, real-time behaviour and compositional safety
We will extend the state-of-the-art modelling frameworks to capture the knowledge learned during evolutions, such as, when and how can a component be reused, performance metrics recorded during the runtime, design rationale for a design decision, time and effort required for different development phases. We will develop non-intrusive analysis and monitoring techniques to systematically collect information during the evolutions, and link this information to the models.
Often, embedded system architectures are derived with little concern for extensibility, rendering evolutions very costly. We will develop decision support methods and tools for the creation of system architectures that are extensible, but without compromising other objectives such as performance, cost, energy consumption, safety and dependability. The evolution of models is often done manually which is tedious and error prone, without any systematic model management. We will develop methods and tools for model management and visualisation, to automate the management of models and improve comprehension during the system evolution.
Table of Contents
Section 1 - Relevance and contributions to the content and objectives of the Call
1.1Relevance
Section 2 - R&D innovation and technical excellence
2.1Concept and objectives
2.2Progress beyond the state-of-the-art
Section 3 - S&T approach and work plan
3.1Quality and effectiveness of the S&T methodology and associated work plan
Section 4 - Market innovation and market impact
4.1Impact
4.2Dissemination and exploitation
4.3Contribution to standards and regulations
4.4Management of intellectual property
Section 5 - Quality of consortium and management
5.1Management structure and procedures
5.2Individual participants
5.3Consortium as a whole
5.4Resources to be committed
Annex A – Funding calculation forms
Section 1 - Relevance and contributions to the content and objectives of the Call
1.1Relevance
Today, most systems are engineered in an evolutionary fashion, but none of the current model-based engineering solutions support evolutionary development. The objective of DECISIVE is to develop a methodology and tool support for model-driven evolutionary design of complex embedded systems.
DECISIVE will contribute to the following industrial priorities of the AWP 2011:
Design methods and tools: DECISIVE will provide model-based engineering methods and tools to overcome the current problems with evolutionary development by: (1) capturing in the models the knowledge gained by developing the previous product versions, (2) increase model reuse by improving the maintainability and comprehension of models (3) provide tool support for the design of extensible system architectures, and (4) support decision makers by improving the accuracy of information used to take decisions in the early stages.
Reference designs and architectures: A further goal of DECISIVE is to provide resource-efficient, non-intrusive, and deterministic monitoring techniques to extract properties from executing systems during both test and deployment.
DECISIVE will contribute to the priorities of the following sub-programmes:
ASP1: Methods and processes for safety-relevant embedded systems: We aim at major technological breakthroughs in the areas of Requirement Management, and Architecture Modelling and Exploration, which will contribute to progress in the areas of Design for Reuse and Design for Safety.
•DECISIVE will contribute to a European Standard Reference Technology platform, proposing meta-models, methods and tools for evolutionary development. The focus on evolutionary development will reduce the times needed for re-certification and re-qualification after change.
•There is a lot of knowledge gained from building a previous version of a product, and from the design flow used. Currently, this information is used informally, and is not systematically captured in the models. No solutions exist for back-annotation of information from previous product versions, gained over the whole previous development cycle: from simulations, analysis, runtime monitoring, testing, etc. We will propose (meta-) models dealing with model evolution and reuse.
•Model reuse can be supported by improved model comprehension and model management. We will propose (semi-) automated model transformations to support evolutionary development. Models will have to be updated based on the knowledge learned in the previous product development. In evolutionary development, handovers between teams could be significantly improved by automated generation of standardized model-views and semi-automated editing that prevent errors from being entered into the model.
•“Architectural design decisions are largely based on experience of past designs and this is difficult to apply to new situations” (ARTEMIS Strategic Research Agenda). The accuracy of decisions will be improved through the use of information gained from previous product versions. We will develop decision support methods and tools for the synthesis of system architectures that are extensible, increase quality of the product, reducing the time and development cost of evolutions.
ASP5: Computing platforms for embedded systems: There is a strong connection between the contributions of the DECISIVE project to ASP1 and ASP5. To support evolutionary design, we have to record information during the runtime of previous product versions. The information recorded in this way will be fed back to the high-level models to be used in designing the new product versions. The SRA identifies the challenge of “evolvability”. We will propose architectural design patterns that improve evolvability. DECISIVE will provide methods and tools that will allow trade-offs between evolvability and other properties such as cost and performance.
DECISIVE addresses the following overall ARTEMIS targets:
- Reduce the cost of system design by reusing models and knowledge from previous products.
- Achieve reduction in development cycles through decision support and trade-off analysis tools.
- Manage complexity increase by facilitating the reuse of models from previous product versions and by (semi-) automatic model management to aid comprehension across evolutions.
- Reduce the effort required for re-validation and re-certification by capturing and reusing the knowledge gained in the validation and certification of the previous product versions and by the reuse of artefacts for qualification and certification.
Section 2 - R&D innovation and technical excellence
2.1Concept and objectives
The main goal of DECISIVE is to develop a methodology and tool support for evolutionary development of complex embedded systems using a model-based design flow. Based on modelling languages that support both system design and analysis the project develops technology for annotation of system models with the results from verification and validation (V&V) activities and the empirical knowledge of the properties of the system and its components. These annotations are to be used to guide decisions both for product-management and for technical development during systems evolution, the annotation are also useful artefacts to support certification and qualification of safety related functions.
The motivation for the project is that contemporary technologies for model-based engineering do not account for the fact that most systems are developed in an evolutionary fashion. Contemporary technology lack functionality and needs improvements in the following aspects:
- Methods to systematically and homogenously associate knowledge of system properties obtained during various V&V activities and from deployed systems to the models used at various design stages and check their consistency. That is, current modelling techniques focus on describing what we want the system to do (as-specified), but not what it actually does (as-is).
- Low- and non-intrusive monitoring methodologies with architecture independent APIs for system analysis, with the purpose both to extract run-time properties to be back-propagated to the model and to monitor and validate the system functionality.
- Efficient and effective model and code management. Understanding and modification of models is often hindered by ad-hoc modelling techniques, bulky graphical presentations, non-intuitive model editors, and lack of adequate two-way synchronization between models and code.
- Support for decision making based on quantitative and qualitative information about the product. That is, product management and architects have no effective tools or models to aid steering the product requirement and design during development projects and between releases.
To bridge these gaps DECISIVE will create innovative solutions by addressing the following objectives:
•Extend existing modelling languages to support annotation of system properties obtained during various V&V-activities (such as analysis, simulation, testing, and safety certification). These extensions will support product-line variability and versioning to fit the evolutionary system development process.
•The modelling tools will be associated to V&V techniques to support consistency checking of back-annotations and the analysis and simulation of models at various abstraction levels, e.g., where part of the system exists in its final form and part of the system only exist as abstract models. This will allow a “what-if” analysis to an early-stage evaluation of the impact of proposed evolutionary changes with respect to, quality attributes (e.g., safety).
•Develop model-to-model transformations and APIs that allow tracing of system properties through design stages and compilation phases. Also, transformations and APIs that allow mapping V&V-results back to the model needs to be provided.
•Ensure availability of information from early verification and validation techniques, allowing business and architectural decisions to be based on solid technical data throughout the development process. Methods for defining and visualizing data from models and prototypes will improve the process for decision makers.
•Improve model-editing capabilities and easy model understanding. In evolutionary system development, handovers between teams will be significantly improved by automated generation of standardized model-views and semi-automated editing preventing errors from being introduced into the model. Also, in the context of model-to-model transformations, and code-to model synchronization, automated model-views will improve readability.
•Implement resource-efficient, non-intrusive, and deterministic monitoring techniques to extract properties from executing systems during both test and deployment.
Project results
The project delivers models methods and tools for evolutionary model based development of embedded systems. It addresses industrial applicability, involving different ways of working, the knowledge and experience of people, and the today's incompleteness of existing tools. The project results will be applied and validated in a diversity of domains that do evolutionary product development with the following characteristics:
•Longevity– Systems that have to be operationally for many years or decades will be updated regularly during maintenance activities. Many version of the same system are operational, and evolutionary development has to take this diversity into account. For these developments, the project has to address the high quality that is required, the necessary evolution of the product and platform that have to be available over long times,and low-cost maintenance over the whole life time of such systems. This leads to the societal benefits of the fast development of new products via evolution, and for high utilization of systems through their long uptime
•Tuning & Scaling– Embedded systems always have a hardware component that has to be operated by the software. Physical variations in production process of hardware leads to variability that has to be incorporated during initialisation and run-time.This presupposes efficient calibration of the system. In addition embedded systems can be highly configurable because of user and environmental variability. The evolution of these systems has to ensure that system variants can still evolve from lessons learned from other variants. The industrial partners in the project that deploy these results deliver societal benefits in reduced fuel consumption and low CO2 emission of cars, and reduced material use and cost during production and enhanced security of other products.
•Reliability Safety– Embedded safety critical systems are expected to work reliable and safe. To address these issues the development has to be conform existing safety and reliability standards, and in addition at run-time, measurements have to be done to evaluate the reliability and safety aspects. The evolution of the systems has to deal with safety and reliability measures, both in the evolution of these aspects themselves, and ensuring that evolution is not hampered by the deigns for reliability and safety designs. An important requirement is that the design should be compositional for these aspects. This addresses the societal need of predictable security and safety and norm compliance to safety standards.
2.2Progress beyond the state-of-the-art
The objective of DECISIVE is to deliver models, methods and an integrated tool chain consolidating novel technology for evolutionary software development and run-time platform support for monitoring. As such a tool chain is not yet available; the project will set a new standard in this field.
DECISIVE will extend state-of-the-art with respect to the above identified aspects as follows:
- We have often witnessed a conflict, or at least non‐consistent, data during the life cycle of a product (e.g. the longest response‐time for an event could be quite different when obtained through scheduling analysis, simulation, testing or observation of the final system).
Existing modelling standard languages, such as MARTE and SysML, lack a structure that allows storage of empirical properties which are attained at different stages during the development process, and with varying degree of quality and confidence.
Also, modelling languages that support product-line variability, such as EAST-ADL2, lack possibility to trace system properties through the variation points. The DECISIVE project will add support to model which properties are preserved over a variation point (and conversely, which properties are affected by a variation).
For early impact analysis, contemporary techniques lack the support to analyse subsystems of various abstraction levels. E.g. scheduling analysis and execution-time analysis typically depend on clock-cycle accurate representation of the final executable, whereas high-level analysis with e.g. Petri-nets and time-automata cannot accurately and safely model execution characteristics of modern hardware platforms. Thus, hybrid techniques are called for.
- Existing model-to-model, and model-to-code, transformations focus on semantic preservation, and sometimes, also property-preservation (e.g. the CHESS modelling language, currently developed in an ARTEMIS-project ending in 2011). However, there is no support in the transformations that allow back-tracing of properties from the target to the source. E.g. there are no methods that allow the memory associated to a signal-queue to be traced back to a particular connection between two components, or more complex, to associate the execution time of a task to the response-time of a signal path in the model. In order to obtain such traceability, we need both to associate meta-data to the target-models, indicating their sources, and provide for structured and automated insertions of probes in the target to allow tracing of interesting properties. Furthermore, processing of the probe-data in order to extract relevant properties and relate them back to the model needs new model-guided analysis technologies and text-to-model transformation techniques.
While most operating systems for embedded and real-time systems provide some performance monitoring mechanisms, e.g. supporting memory-profiling and task-level execution monitoring, these mechanisms do not give detailed enough data to allow back-annotation of properties to models. Conversely, naïve instrumentation of code during model-to-code transformation will likely consume too many resources in terms of both execution time and memory. To remedy this situation, we will develop platform-level monitoring techniques both in hardware for non-intrusive monitoring, and in software for low-intrusive monitoring, that can be automatically customized with respect to the amount of resources required. These mechanisms can then be used by the model-to-code transformations. We will also implement optimisation techniques to limit both the amount of probes that need to be generated and the amount of data that need to be stored for each probe in order to obtain a given quality of the observation.