visibly linear dynamic logic
play

Visibly Linear Dynamic Logic Joint work with Alexander Weinert - PowerPoint PPT Presentation

Visibly Linear Dynamic Logic Joint work with Alexander Weinert (Saarland University) Martin Zimmermann Saarland University December 14th, 2016 FSTTCS 2016, Chennai, India Martin Zimmermann Saarland University Visibly Linear Dynamic Logic


  1. Visibly Linear Dynamic Logic Joint work with Alexander Weinert (Saarland University) Martin Zimmermann Saarland University December 14th, 2016 FSTTCS 2016, Chennai, India Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 1/17

  2. The Everlasting Quest for Expressiveness Consider an arbiter granting access to a shared resource. Requirements: “Every request q is eventually answered by a response p ” “Every request q is eventually answered by a response p after an even number of steps” “There are never more responses than requests” Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 2/17

  3. The Everlasting Quest for Expressiveness Consider an arbiter granting access to a shared resource. Requirements: “Every request q is eventually answered by a response p ” Linear Temporal Logic: G ( q → F p ) “Every request q is eventually answered by a response p after an even number of steps” “There are never more responses than requests” Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 2/17

  4. The Everlasting Quest for Expressiveness Consider an arbiter granting access to a shared resource. Requirements: “Every request q is eventually answered by a response p ” Linear Temporal Logic: G ( q → F p ) “Every request q is eventually answered by a response p after an even number of steps” Linear Dynamic Logic: [ true ∗ ]( q → � ( true · true ) ∗ � p ) “There are never more responses than requests” Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 2/17

  5. The Everlasting Quest for Expressiveness Consider an arbiter granting access to a shared resource. Requirements: “Every request q is eventually answered by a response p ” Linear Temporal Logic: G ( q → F p ) “Every request q is eventually answered by a response p after an even number of steps” Linear Dynamic Logic: [ true ∗ ]( q → � ( true · true ) ∗ � p ) “There are never more responses than requests” Expressible with pushdown automata/context-free grammars as guards ⇒ Visibly Linear Dynamic Logic Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 2/17

  6. Outline 1. Preliminaries 2. Expressiveness 3. VLDL Verification 4. Discussion Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 3/17

  7. Outline 1. Preliminaries 2. Expressiveness 3. VLDL Verification 4. Discussion Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 4/17

  8. Visibly Pushdown Automata Partition input alphabet Σ into Σ c (calls), Σ r (returns), and Σ ℓ (local actions). A visibly pushdown automaton (VPA) has to push when processing a call, pop when processing a return, and leave the stack unchanged when processing a local action. Stack height determined by input word ⇒ closure under union, intersection, and complement. Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 5/17

  9. Visibly Pushdown Automata Partition input alphabet Σ into Σ c (calls), Σ r (returns), and Σ ℓ (local actions). A visibly pushdown automaton (VPA) has to push when processing a call, pop when processing a return, and leave the stack unchanged when processing a local action. Stack height determined by input word ⇒ closure under union, intersection, and complement. Examples: a n b n is a VPL, if a is a call and b a return. ww R is not a VPL. Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 5/17

  10. Visibly Linear Dynamic Logic (VLDL) Syntax ϕ := p | ¬ ϕ | ϕ ∧ ϕ | ϕ ∨ ϕ | � A � ϕ | [ A ] ϕ where p ∈ P ranges over atomic propositions and A ranges over VPA’s. All VPA’s have the same partition of 2 P into calls, returns, and local actions. Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 6/17

  11. Visibly Linear Dynamic Logic (VLDL) Syntax ϕ := p | ¬ ϕ | ϕ ∧ ϕ | ϕ ∨ ϕ | � A � ϕ | [ A ] ϕ where p ∈ P ranges over atomic propositions and A ranges over VPA’s. All VPA’s have the same partition of 2 P into calls, returns, and local actions. Semantics: ( w ∈ (2 P ) ω ) w | = � A � ϕ if there exists an n such that w 0 · · · w n − 1 is accepted by A and w n w n +1 w n +2 · · · | = ϕ . w | = [ A ] ϕ if for every n s.t. w 0 · · · w n − 1 is accepted by A we have w n w n +1 w n +2 · · · | = ϕ . Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 6/17

  12. Example “Every request q is eventually answered by a response p and there are never more responses than requests” [ A ∗ ]( q → � A ∗ � p ) ∧ ¬� A � true where A ∗ accepts every word, and A accepts those words with more responses than requests. Both languages are visibly pushdown, if { q } is a call, { p } is a return, and ∅ and { p , q } are local actions. Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 7/17

  13. Outline 1. Preliminaries 2. Expressiveness 3. VLDL Verification 4. Discussion Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 8/17

  14. Expressiveness Lemma VLDL and non-deterministic ω -VPA are expressively equivalent. Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 9/17

  15. Expressiveness Lemma VLDL and non-deterministic ω -VPA are expressively equivalent. Proof Idea VLDL non-deterministic ω -VPA Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 9/17

  16. Expressiveness Lemma VLDL and non-deterministic ω -VPA are expressively equivalent. Proof Idea VLDL Deterministic Stair Automata O (2 n ) [LMS ’04] non-deterministic ω -VPA Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 9/17

  17. Expressiveness Lemma VLDL and non-deterministic ω -VPA are expressively equivalent. Proof Idea VLDL O ( n 2 ) Deterministic Stair Automata O (2 n ) [LMS ’04] non-deterministic ω -VPA Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 9/17

  18. Expressiveness Lemma VLDL and non-deterministic ω -VPA are expressively equivalent. Proof Idea VLDL O ( n 2 ) O ( n 2 ) 1-way Alternating Deterministic Jumping Automata Stair Automata O (2 n ) [LMS ’04] non-deterministic ω -VPA Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 9/17

  19. Expressiveness Lemma VLDL and non-deterministic ω -VPA are expressively equivalent. Proof Idea VLDL O ( n 2 ) O ( n 2 ) 1-way Alternating Deterministic Jumping Automata Stair Automata O (2 n ) O (2 n ) [LMS ’04] [Bozelli ’07] non-deterministic ω -VPA Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 9/17

  20. Expressiveness Lemma VLDL and non-deterministic ω -VPA are expressively equivalent. Proof Idea VLDL O ( n 2 ) O ( n 2 ) 1-way Alternating Deterministic Jumping Automata Stair Automata O (2 n ) O (2 n ) [LMS ’04] [Bozelli ’07] non-deterministic ω -VPA Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 9/17

  21. From Stair Automata to VLDL 1 1 2 0 2 7 3 2 2 0 3 0 5 3 0 c ω c c c c r c r r c c c r r l Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 10/17

  22. From Stair Automata to VLDL 1 1 2 0 2 7 3 2 2 0 3 0 5 3 0 c ω c c c c r c r r c c c r r l Acceptance: maximal priority occuring at infinitely many steps even Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 10/17

  23. From Stair Automata to VLDL 1 1 2 0 2 7 3 2 2 0 3 0 5 3 0 c ω c c c c r c r r c c c r r l Acceptance: maximal priority occuring at infinitely many steps even Equivalently: For some state q of even priority c there is step with state q s.t. 1. after this step, no larger priority appears at a step, and 2. for every step with state q , there is a later one with state q . Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 10/17

  24. From Stair Automata to VLDL 1 1 2 0 2 7 3 2 2 0 3 0 5 3 0 c ω c c c c r c r r c c c r r l Acceptance: maximal priority occuring at infinitely many steps even Equivalently: For some state q of even priority c there is step with state q s.t. 1. after this step, no larger priority appears at a step, and 2. for every step with state q , there is a later one with state q .   � �  ∧ [ A ′ � q I A ′ q � [ q A ′ q ] � q A ′ q � true q ′ ] false  q ∈ Q even q ′ ∈ Q > Ω( q ) Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 10/17

  25. Outline 1. Preliminaries 2. Expressiveness 3. VLDL Verification 4. Discussion Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 11/17

  26. Satisfiability Theorem VLDL Satisfiability is ExpTime -complete. Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 12/17

  27. Satisfiability Theorem VLDL Satisfiability is ExpTime -complete. Proof Sketch Membership: Construct equivalent ω -VPA and check it for emptiness. Hardness: Adapt ExpTime -hardness proof of LTL model-checking of pushdown systems [BEM ’97] Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 12/17

  28. Model Checking Theorem VLDL model checking of visibly pushdown systems is ExpTime -complete. Martin Zimmermann Saarland University Visibly Linear Dynamic Logic 13/17

Recommend


More recommend