1 / 15

Model-Based Run-Time Error Detection

system. system output. input. compare model and system (output & state). recovery. error. diagnosis. system state. correction. awareness. Stateflow model of desired behaviour. Jozef Hooman & Teun Hendriks Embedded Systems Institute Eindhoven, The Netherlands.

bruno
Download Presentation

Model-Based Run-Time Error Detection

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. system system output input compare model and system (output & state) recovery error diagnosis system state correction awareness Stateflow model of desired behaviour Jozef Hooman & Teun Hendriks Embedded Systems Institute Eindhoven, The Netherlands Model-Based Run-Time Error Detection Models@run.time, 2 October 2007

  2. TRADER System Reliability Consumer Electronics Research Period: Sept. 2004 - Aug. 2009 20 fte/yr, 7 PhDs, 1 Postdoc, 10 Partners Previously known as Philips Semiconductors Carrying Industrial Partner • Goal • Develop methods and tools to optimize reliability of high-volume products • Issues • Minimize product failures. • Improve user-perceived reliability • Case studies from TV domain Jozef Hooman

  3. Reliability threats in TV domain Increasing complexity • Functions/content increases rapidly • Play music (mp3, ..), view photos, search teletext, Electronic Programming Guide, child lock, sleep timer, Picture-in-Picture, TV ratings, emergency alerts, many image processing options and user settings, … • External information sources multiply • Connected planet strategy, downloadable applications  Increase of SW (1KB in 1980 – 64MB in 2007) Increase of third party content (EPG, codec’s) Decreasing time-to-market • Fixed shipping gates to occupy reserved shelf space Faults in delivered products are a fact-of-life Jozef Hooman

  4. Not satisfying the high reliability expectations Many returned products Damages brand image Reduces market share Cost of non-quality (CoNQ) 2-3% turnover (compare to research expenditures: 2.3%)Rudy Provoost (CEO Philips Consumer Electronics) Challenge: Prevent product faults causing customer complaints given constraints: Low costs Short time to market Business impact Jozef Hooman

  5. Took TV that was on market for half a year Installed latest upgrades Found 20 failures in 20 hours playing with TV 8 public upgrades in few months Example User sees problem immediately, but TV seems unaware of it Jozef Hooman

  6. system system output input compare model and system (output & state) recovery error diagnosis Approach: run-time awareness system state correction awareness Stateflow model of desired behaviour Jozef Hooman

  7. Modeling Which part of the system to model? What are most suitable models, at which level of abstraction? Which models can be implemented most efficiently? How to obtain suitable models? How to increase confidence in the model? system system output input system state correction awareness Stateflow model of desired behaviour compare model and system (output & state) recovery error diagnosis Research Questions (1) Jozef Hooman

  8. First model desired user-perceived behaviour (in Matlab/Simulink/Stateflow) More difficult than expected; not explicit in current requirements documents Model desired behaviour Jozef Hooman

  9. Models at run-time How to preserve model semantics in implementation? How to avoid false errors? When to compare system observations with model? When to report an error, how much tolerance is allowed? system system output input system state correction awareness compare model and system (output & state) recovery Stateflow model of desired behaviour error diagnosis Research Questions (2) Jozef Hooman

  10. Experiment with awareness concept: Simulation using Simulink/Stateflow compare model of System Under Observation (SUO) and SPEC, trydifferent error detection strategies Linux-based awareness framework in which SUO and SPEC can be inserted easily Open source media player MPlayer [current work] model desired behaviour and insert system and spec in Linux-based framework Add awareness to part of TV system [future work] Awareness Experiments Jozef Hooman

  11. Simulating Awareness Comparison can be • Time-triggered (compare pulse) • Event-triggered (compare signal from model) Jozef Hooman

  12. Simulating Awareness Observe internal variable to detect error earlier and allow recovery before user observes failure Jozef Hooman

  13. Application: MPlayer Design of Framework in Linux Jozef Hooman

  14. Stateflow suitable modeling language Allows various forms of model validation Various options for code generation Linux-based frameworks suitable for experiments Comparison strategy needs further study When to compare to avoid false errors In Linux framework we allow the specification of Maximal allowed deviation on values Maximal number of allowed consecutive erroneous deviations Uncertainty about system behaviour, such as worst-case & best-case execution times, are included in model Avoid comparison when systen behaviour is uncertain Concluding Remarks Jozef Hooman

  15. Thank you for your attention! Jozef Hooman

More Related