1 / 19

Possibilistic and probabilistic abstraction-based model checking

Possibilistic and probabilistic abstraction-based model checking. Michael Huth Computing Imperial College London, United Kingdom. Outline of talk. need for abstraction modal quantitative systems possibilistic semantics probabilistic semantics specification of abstractions

alpha
Download Presentation

Possibilistic and probabilistic abstraction-based model checking

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. Possibilistic and probabilistic abstraction-based model checking Michael Huth Computing Imperial College London, United Kingdom

  2. Outline of talk • need for abstraction • modal quantitative systems • possibilistic semantics • probabilistic semantics • specification of abstractions • conclusions.

  3. Need for abstraction LTL model checking for finite-state Markov decision processes is [Courcoubetis & Yannakakis’95] • polymonial in model (which are big) and • doublyexponential in formula. Infinite-state models occur in practice.  Aggressive abstraction techniques required for model checking real-world designs.

  4. Abstraction loci Abstract the computation of a model check M |=f, by approximating • the model M to M*; e.g. simulations [Larsen & Skou’91] • the satisfactionrelation|= to |=*, e.gcompositionalconjunction [Baier et al.’00] • the propertyf to f*, e.g. bounded model checking [Clarke et al.’01] Combinations possible: e.g. make a probabilistic M non-probabilistic [Vardi’85].

  5. Soudness needed • Valid verfication certificates: positive abstract check M* |=* f*  M |=f holds as well. • Valid refutationcertificates: nevative abstract check M* |=* ¬f* M |= ¬ fholds, too. • Rangeoff : full logic forsound mix of fairness & abstraction, safety & liveness, verification & refutation, etc. Such a framework is well developed for qualitative systems: three-valued model checking [Larsen & Thomsen’88, Bruns & Godefroid’99].

  6. Research aims • transfer two-valued & three-valued model checking toquantitativesystems; • let probabilisticsystems be a specialinstance of such a transfer; and • use transferred results to re-assessexistingwork on abstraction of probabilistic systems.

  7. Modal quantitative systems • modal nature of non-determinism: “There are delays on the Bakerloo Line.” != “There are no delays on the remaining lines.” • transitions (s,m) have type S x [F P] - P partial order of quantities - Fs-algebra on state set S - [F P] = maps m :F P such that A in A’  m(A) £ m(A’) • atomic observables and preimage operator are in F.

  8. Examples • “neural” systems - each s in S is a stimulus ws in [0,w) - m(A) is weighted sum of stimuli ws • Markov decision processes - P = [0,1] - all mintransitionsare probability measures - complete: non-determinism fully specified • Choquet’scapacities,pCTL*, andweakbisimulation [Desharnais et al.’02].

  9. Concrete and abstract model gQis special s0 t0 = { s0, s1, s3 } .5 a .25 p .5 p? q? .25 hQ 1/3 m 2/3 q .5 2/3 .75 1/3 s1 1/3 1/3 aQ mQ 1/3 g .5 .25 h pq 2/3 .5 gQ s3 s2 1/3 p @ (p = tt) is valid t1 = { s2 } p? @ (p = tt) is satisfiable

  10. Measurable navigation • a relation Q : S1 S2has measurable navigation: for all A in F1 and Bin F2 A.Q in F2and Q.B in F1 • non-trivial property • basis for relational abstraction/refinement • works for finite quotients with measurable equivalence classes.

  11. Lifting relations to measures For Q : S Swith measurable navigation, define Qps : [F  P]  [F  P] by (m,h) in Qps iff for all A, B in F m(A) £h(A.Q) andh(B) £ m(Q.B)  … a generalization of probabilistic (bi)simulation [Larsen & Skou’91].

  12. Abstraction & refinement A relation Q : S S with measurable navigation is a possibilisticrefinement if (s,t) in Q implies • (t,h) in Ra  $(s,m) in Ra such that (m,h) in Qps • (s,m) in Rc  $(t,h) in Rc such that (m,h) in Qps Ra = guaranteed transitions (e.g. gQabove), Rc = possible transitions. //modal non-determinism

  13. Possibilistic semantics Quantitative logic: • f::=tt | p | Z | mZ.f | ¬f | f & f | EX>r f • assertion checks s|=a f • consistency checks s|=c f • usualsemantics, exceptfor - s|=a ¬ fiff not s|=cf; - s|=c ¬ f iff not s|=a f; and - s|=l EX>r f iff $(s,m) in Rl : m({t | t|=l f}) > r where l in {a, c}.

  14. Soundness Weprove { s in S | s|=l f} in F for l in {a, c} andfand use it to show: “Q possibilisticrefinementwith (s,t) in Q, then 1. t|=a f s|=a f 2. s|=c f t|=c f // needed to prove 1. for all f.”

  15. Probabilistic semantics • probability measures for transitions • mZ.f restricted to probabilistic EU • same semantics except for EU • possibilistic semantics “approximates” probabilistic one • sound probabilistic refinement: Q  Qpr [Larsen& Skou’91] • Qpr =Qps forfinite-stateMarkovdecisionprocesses.

  16. Specification of abstraction S = state set of un-abstracted model, A = finite target state set of abstract model: • specify left/right-total relation Q : S A; • determines an abstract model over A with discrete s-algebra … • … which makes Q into a refinement.

  17. Understanding the lift • m in [F P]  mQ (B) = m(B.Q) welldefined • (m,mQ) in Qps • (m,h) in Qps  h£mQ • converse of 3. holds if Q is graph of a function • S finite state set of Markov decision process Qps = Qpr & sameabstractions … 4. holdsif A isafinitesetofmeasurableequivalenceclasses, e.g. predicateabstraction w.r.t. finitely many measurable predicates.

  18. Example re-visited s0 t0 = { s0, s1, s3 } |=a ¬EX >3/4 ¬EX>3/10¬p .5 a .25 p .5 p? q? .25 hQ 1/3 m 2/3 q .5 2/3 .75 1/3 s1 1/3 1/3 aQ mQ 1/3 g .5 .25 h pq 2/3 .5 gQ s3 s2 1/3 Abstraction along the predicate ¬(¬p & ¬q) t1 = { s2 } only gQin Ra

  19. Conclusions • transferred three-valued model checking toquantitativesystems; • showedthat probabilisticsystems and Larsen & Skou simulations are a specialinstance of such a transfer; • re-assessedexistingwork on abstraction of probabilistic systems in this context; and • showed that this approach works for an important class of finite-state abstractions.

More Related