early time budgeting in distributed embedded control
play

Early Time-Budgeting in Distributed Embedded Control Systems Manoj - PowerPoint PPT Presentation

Early Time-Budgeting in Distributed Embedded Control Systems Manoj noj G. Dixi xit, t, Jointly with, Prof. . Palla lab Dasgupt pta, , IIT KGP and Dr. S. Ramesh, h, GM R&D CMI Workshop: Making Formal Verification Scalable and Usable


  1. Early Time-Budgeting in Distributed Embedded Control Systems Manoj noj G. Dixi xit, t, Jointly with, Prof. . Palla lab Dasgupt pta, , IIT KGP and Dr. S. Ramesh, h, GM R&D CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 1

  2. Ba Backgroun ound Moder dern embed bedded ed contro rol l syste tems ms are comp mpon onent ent  based sed and hav ave e larg rge e number mber of features tures  E.g. AUTOSAR based development Safet ety y features tures have e stri rict ct real-ti time me end-to to-en end d  requir quiremen ements Many y compon onen ents ts interac eract t toget ether her to meet t syste tem m  level el requir quirements ements Syst stem em is distrib stributed uted in natur ure  How to do a timing layout of entire system to meet end-to-end real-time requirements? CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 2

  3. Il Illustra ustrativ tive e Exa xample ple Functional Specification Safety Requirement CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 3

  4. illu lustrativ strative e Exa xample mple Functional Specification Standards/Statistical data Safety Requirement Actual Implementation Time Budgeting CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 4

  5. Emerging rging Challe allenges nges Increasing reasing complex mplex features tures  Multi ltipl ple e func nction tions s in n a single e computa putati tional al uni nit, , e.g. . AU AUTOS OSAR R  More re comp mpon onen ent sharing aring prom romot oted ed by the smaller ller compon onen ent size zes  Need ed for advance ance plann nning ing of resou source rces s for extens ensibili lity ty  This is leading to…   Increasing real-time interdependencies between components CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 5

  6. Pr Prev evalent alent Ap Appr proache oaches Ad-ho Ad hoc c estimat imates es about out compon onen ent  resp sponse onse time me Arch chitectur ecture e explora orati tion n to do  comp mpone nent-task task mapping ping, comp mpone nent-ECU CU mappi ping ng etc. On fail On ilur ure, e, diffic fficult t to trace ce the  culp lprit rit comp mpone nent How to budget time for each component is not clear CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 7

  7. Our Pr Prop oposal sal Earl rly y tim ime-budg budget eting ing for embed bedded ded  control rol-system systems Comp mpon onen ent have e parame ametric tric timi ming ng  requir quiremen ements Use forma mal l specifi ecifica cation tion and  analysis ysis method hodolog ology y to gener erat ate e const stra raint nts s on para ramete meter r valua luations tions CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 8

  8. Example Ex mple Real-time Specifications Parametric -time Specifications Time-Budgeting: What values of x, y, w, z are good-enough? CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 9

  9. In Indu dustrial strial relev levanc ance Shi hift ft towar ards s earl rly y speci ecifi fica cation tion of timin ming g  requir quiremen ements Large component integration, multiple suppliers   Important to know, how a specific choice of timing specification for one component affects the other AUTOSAR SAR meta a model del allow ows s speci ecifyi fying ng timin ming g  specifica ecificati tions ns at dif ifferen ferent levels els of softwa tware re Source: Autosar timing spec from http://www.autosar.org hier erar archy chy – comp mpone nents ts to network ork  EAST-ADL and TIMMO2 provides higher level of abstractions for specifying functional and product line requirements  Timing requirements are refined across different levels Event models – periodic, sporadic etc.  Delay, synchronization constraints  Source: TIMMO Methodology presentation by Stephan Kuntz, Continental Automotive GmbH, 2010 CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 10 10

  10. The Pr Prob oblem lem  We are e given ven a set t of of features tures and d th their r real al-time time requ quire irements ments  We are e given ven a set t of of com ompon onents nts and d th their r param ramet etric ric-ti time me requirem irements ents for or impl pleme menting nting th these feat atures ures  Prop opos ose e Early ly sta tage ge Time me-Budge udgeting ting Meth thod odol olog ogy  Find constraints over parameter values Design space exploration to select suitable valuation   Scalable CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 11 11

  11. Sc Scalab ability ility and d Usability bility Consid nsiderations erations -1 In pract ctic ice, , component ponent decomp mposi ositio ions ns are hierarchic ical: l:  Handling ing large ge hierarchical archical decomp mposi ositio ions ns:  DAG  Split component time-budgeting into smaller sub-  Methodology aligned for hierarchical specifications problems and repeat Compositional approach  Large e decompo mpositi ition ons: s:   Each feature component has 10s of requirements  DF-traversal with back-tracking takes care of Simultaneous budgeting does not seem to be  component re-use case scalable  Linear constraints are preferred Specialized lized methods ds to analyze requirement ments s patterns rns  Requir iremen ents ts become me finer er and more comple plex x (or  detaile led) d) as we move down the hierarchy archy CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 12 12

  12. Ti Time-Budgeting Budgeting Sing ngle le Step Feature Component Requirements Requirements (known-timing) (unknown-timing) Formalized Feature Formalized Component Design Optimization Requirements Requirements Constraints Decisions (Real-time) (Parametric-time) Constraint Solving, Optimization: New Algorithms Well known Formal Analyzer of Constraint Linear Constraints Parametric Specs Optimizer On Parameters Time Budgeting Algorithm • Novelty : System level optimization Component Time- gets converted to constraint solving, Budget scalability is much better this way CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 13 13

  13. Formali rmalizati zation n of Requ quireme irement nt Decomposition mposition Requ quirem rements ents decom compositi position n step ep is formal alize zed as a collec ection tion of requir quirement ement decom omposi position tion pai airs rs  f : feature ture requ quire rement ent and let g 1 , …, g k are compon onen ent requ quiremen ements ts identi entified fied for f  (f, { g 1 , …, g k }) }) is a requirem quirement ent decom compos positi tion pair ir  Veri rification fication check: eck:   Informally: component requirements put-together should satisfy all feature requirements  We have the following reduction: Theorem: It is enough to analyze each pair separately, compute validity constraint. Any solution to conjoined constraint defines a suitable time-budget CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 14 14

  14. Formal rmal Sp Specific ification ation for r Requirem irements ents  Parame rametric tric Tempor oral al Log ogic ic (PLTL) TL)  Extends well known Linear Temporal Logic Semantics is defined by using a parameter valuation  is a Requirement Decomposition Pair CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 15 15

  15. Va Valid idity ity of a Requirem irement ent Decomposition mposition Pa Pair The requirement decomposition pair is valid if and only if PLTL formula is valid. Due to the parameters, this reduces to constraint computation CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 16 16

  16. Abstractly… Validity Region x 2 x1 Valid φ x2 Not Valid PLTL Formula x 1 ula φ , we want Give iven n a PLTL TL for ormula t to to f find nd th the representatio sentation n of of th the sol olution tion regi gion on in th the for orm of of a con onstraint traint CMI Workshop: Making Formal Verification Scalable and Usable Manoj Dixit 10 th Jan 2013 17 17

Recommend


More recommend