Model Quality Objectives for Collaboration between OEM and Suppliers

Size: px
Start display at page:

Download "Model Quality Objectives for Collaboration between OEM and Suppliers"

Transcription

1 Model Quality Objectives for Collaboration between OEM and Suppliers Florian Levy (Renault), Stéphane Louvet (Bosch), François Guérin (MathWorks) April 17 th, The MathWorks, Inc.

2 Agenda 1. Background and Motivations 2. Model Quality Objectives and MathWorks support 3. Deployment and expected gains 4. Conclusion

3 Co-development in automotive System and Software co-development between teams and companies leads to increase model sharing. However difficulties are regularly faced. Customer Plant Model Implementation model Prototyping model Reactive model System model Supplier Exchanges / Discussions Functional model SW Model Architecture Model Control model Model Which kind of model is shared and what is its content?

4 Automotive customer/supplier Model Exchange History -2005: model not shared, only specification SW Development based on textual requirements Textual Spec : start to receive models in addition of textual specifications Model aim: for information, to understand specifications and dynamics SW Development based on textual requirements Textual Spec : start to receive some model as requirement reference Model aim: solution to be coded, regularly without explanations, and no test scenarii SW Development based on model? SW methodology needs to be adapted and shared between actors

5 Concrete difficulties faced on automotive model exchanges (1/3) 1 st use case: use a prototyping Model to create a SW-Component Model provided: Model working on prototyping system No txt req., and so no traceability in the model Scenarii provided (not tests) Main difficulties faced: How to ensure conformity with specification at model or code level? How to satisfy Model coverage? Code coverage? (ISO >= ASIL A) ASIL D ASIL C ASIL B ASIL A QM

6 Concrete difficulties faced on automotive model exchanges (2/3) 2 nd use case: use a functional reference Model and integrate the SW-C Model provided: Model is a reference (forbidden to modify it), except for SW implementation No specification, and so no traceability to upstream requirements No test scenarii Main difficulties faced: How to be efficient for automotive design rules conformity (MAAB, MISRA )? Ex: when a MAAB rule is violated, needed to ask customer for modification, then new delivery How to have the same understanding of remaining effort? Ex: often considered by customer as almost ready for coding

7 Concrete difficulties faced on automotive model exchanges (3/3) 3 rd use case: use a large Model and integrate the SW-C Model provided: Very large Model No specification, and so no traceability to upstream requirements No test scenarii Main difficulties faced: How to succeed its integration in the full SW? How to perform full System Tests? How to ensure ISO ASIL compliance or A-SPICE compliance? Ex: freedom from interference, bidirectional traceability in SW Unit and Tests

8 Some additional difficulties with model sharing In addition of these points Who is responsible of what? Who is responsible of the quality of code? What to do when non conformities happen? To support this new SW methodology based on model sharing, all these questions needed answers! Model Quality Objectives (MQO)

9 Agenda 1. Background and Motivations 2. Model Quality Objectives and MathWorks support 3. Deployment and expected gains 4. Conclusion

10 Model Quality Objective working group Idea emerged from SQO success few years ago New working group focused on model (MQO) extended to Bosch Consensus-based decisions ~100h of meetings over 2,5 years many document reviews Jérôme Bouquet Stéphane Faure Florent Fève Matthieu Foucault Ursula Garcia François Guérin Thierry Hubert Pascal Lamothe Florian Levy Stéphane Louvet Patrick Munier Pierre Nicolas Paton Alain Spiewek Philippe Spozio Yves Touzeau Renault Valeo Valeo PSA Bosch MathWorks PSA PSA Renault Bosch MathWorks Delphi Delphi Renault Renault

11 MQO clarifies exchanges and discussions Functional model (MQO1) Customer Architecture model (MQO2) Supplier Exchanges / Discussions Component Design model (MQO3) MQO Component Implementation model (MQO4)

12 Types of Models

13 Model Quality Objectives Design model name Functional model Architecture model Component design model Component implementation model Quality Objective MQO-1 MQO-2 MQO-3 MQO-4 Set of Model Quality Requirements to be able to assess the quality of each type of models MQR ID MQR Title MQO-1 MQO-2 MQO-3 MQO-4 MQR-01 Model layout M M M M MQR-02 Model comments M M M M MQR-03 Model links to requirements M M M M MQR-04 Model testing against requirements M R M M MQR-05 Model compliance with modeling standard M M M MQR-06 Model data M M M MQR-07 Model size M M MQR-08 Model complexity M M MQR-09 Model coverage M M MQR-10 Model robustness M M MQR-11 Generated code testing against requirements R M MQR-12 Generated code compliance with coding standard R M MQR-13 Generated code coverage R M MQR-14 Generated code robustness R M MQR-15 Generated code execution time M MQR-16 Generated code memory footprint M

14 Example of Model Quality Requirement Example of a Model Quality Requirement

15 Guidelines on model reuse

16 MathWorks Metric support All MQR can be measured with MathWorks tools Simulink Check provides additional metrics for size, architecture, compliance and readability Metric dashboard introduced in R2017b is rapidly involving to display and navigate from metrics results to models navigate to model block

17 Agenda 1. Background and Motivations 2. Model Quality Objectives and MathWorks support 3. Deployment and expected gains 4. Conclusion

18 Communication Internal company communication Communication to engineering teams for feedback (engineering, quality, safety) Presentation to management for approval Public communication ERTSS 2018 (February 2018) MathWorks Automotive Conference Stuttgart (April 2018) MATLAB Expo (June 2018) MathWorks website (2018)

19 Example of deployment at Renault / Valeo Training MQO will be part of some standard MBD training for new users. (Renault/Valeo) Large internal communication on MQO is planned for the coming months. (Renault/Valeo) Projects MQO will be integrated in a future version of our Request For Quotation package as an additional reference. (Renault) Process MQO will be taken into account as an input for our future software development process. (Renault)

20 Expected gains The organizations that apply MQO should experience the following benefits: Shared understanding of Model-Based Design within the organization Application of a quality model adapted to MBD projects and compatible with industry software quality and safety standards Assessment of model quality at different phases of projects The organizations that also collaborate with partners to execute MBD projects should experience the following benefits: Clear split of responsibility between parties at the beginning of projects Common understanding of model quality Common expectation on model quality when sharing models

21 Agenda 1. Background and Motivations 2. Model Quality Objectives and MathWorks support 3. Deployment and expected gains 4. Conclusion

22 Conclusion We expect MQO to improve MBD co-development between customers and suppliers. We look for feedback to improve MQO: Additional Model Quality Requirements? Additional types of models (e.g. system level)? Applicability in other industries?