a canonical model construction for iteration free pdl
play

A Canonical Model Construction for Iteration-Free PDL with - PowerPoint PPT Presentation

A Canonical Model Construction for Iteration-Free PDL with Intersection Florian Bruse Daniel Kernberger Martin Lange University of Kassel, Germany September 22, 2016 Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0


  1. A Canonical Model Construction for Iteration-Free PDL with Intersection Florian Bruse Daniel Kernberger Martin Lange University of Kassel, Germany September 22, 2016

  2. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Canonical Models Tool to show completeness of proof calculus (for e.g., ML) Idea: • take set of maximally consistent sets of formulas (mcs) as underlying set of structure • atomic propositions via membership a • Φ − − → Ψ iff [ a ] ¬ ψ ∈ Φ for no ψ ∈ Ψ a [ a ] ¬ ψ ∉ ∋ ψ Φ Ψ → (via induction): ϕ true at Φ iff ϕ ∈ Φ. yields satisfiability of any consistent set of formulas, i.e., completeness. NB: presence of edge depends only on endpoints. 2 / 9

  3. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Iteration-Free PDL with Intersection (PDL 0 ) fix propositions { P , Q ,... } = P , atomic programs { a , b ,... } = R Syntax: formulas: ϕ ∶∶ = P ∣ ϕ ∨ ϕ ∣ ϕ ∧ ϕ ∣ ¬ ϕ ∣ ⟨ α ⟩ ϕ ∣ [ α ] ϕ programs: α ∶∶ = a ∣ α ; α ∣ α ∩ α ∣ α ∪ α ∣ ϕ ? Semantics (sketch) over LTS T : • ⟨ α ⟩ ϕ true at s iff ex. t with s − − α → t and ϕ true at t a → t iff ( s , t ) ∈ a T • s − − α 1 ; α 2 • s α 1 α 2 − − − − − − → t iff ex. u with s − − − → u and u − − − → t α 1 ∩ α 2 α 1 α 2 • s − − − − − − → t iff s − − − → t and s − − − → t ϕ ? • s − − − → t iff s = t and ϕ true at s 3 / 9

  4. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 PDL 0 in action a ⊧ P ∧ ⟨ a ⟩ P ∧ [ a ∩ ⊺ ? ]� P P → no tree model property b a P , Q ⊧ ⟨ a ; [ b ; P ? ]¬ Q ?; b ⟩( P ∧ Q ) / → convoluted and nested programs hard to conquer inductively 4 / 9

  5. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 More complications Consider sat. set Φ (e.g. theory of dead end world) Ψ = ⋃ ϕ ∈ Φ {⟨ a ⟩ ϕ, [ a ] ϕ, ⟨ b ⟩ ϕ, [ b ] ϕ } ∪ {[ a ∩ b ]�} Ψ has model: b a Φ Ψ Φ But no model with only one instance of Φ → canonical model needs adaption Existing constructions not convincing enough 5 / 9

  6. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 A Proof Calculus for PDL 0 Standard style proof system with derivation rules ( MP ) ϕ ϕ → ψ ϕ ϕ ( Gen ) ( USub ) ψ [ α ] ϕ ϕ � ψ / p � α ⇒ α ′ ( PSub ) ϕ ϕ �⟨ α ′ ⟩/⟨ α ⟩� ( pos ) and axioms and axiom schemes: α ∩ β ⇒ α ( p ?; α ) ∩ β ⇔ p ?; ( α ∩ β ) ... 6 / 9

  7. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model Idea: build “free” structure, i.e., maximally tree-like, no unnecessary connections 7 / 9

  8. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Φ Φ ′ start with mcs, no edges → atomic and box formulas satisfied (generation 0) 7 / 9

  9. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Φ Φ ′ 7 / 9

  10. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α Ψ ,α Φ Ψ ,α ′ Φ ′ add witnesses for missing diamonds 7 / 9

  11. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α β Ψ ,α Φ α ′ Ψ ,α ′ Φ ′ add witnesses for missing diamonds, connect with abstract edges 7 / 9

  12. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α β Ψ ,α Φ α ′ Ψ ,α ′ Φ ′ add witnesses for missing diamonds, connect with abstract edges in disjoint fashion 7 / 9

  13. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α β Ψ ,α Φ α ′ Ψ ,α ′ Φ ′ 7 / 9

  14. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α β Ψ ,α Φ α ′ Ψ ,α ′ Φ ′ refine iteratively α = α 1 ∩ α 2 7 / 9

  15. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α 1 β Ψ ,α Φ α 2 α ′ Ψ ,α ′ Φ ′ refine iteratively α = α 1 ∩ α 2 7 / 9

  16. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α 1 β Ψ ,α Φ α 2 α ′ Ψ ,α ′ Φ ′ refine iteratively α = α 1 ∩ α 2 7 / 9

  17. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α 1 β Ψ ,α Φ α 2 α ′ Ψ ,α ′ Φ ′ refine iteratively, add intermediate nodes if necessary α = α 1 ∩ α 2 β = β 1 ; β 2 7 / 9

  18. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α 1 β 2 Ψ ,α Ψ ′′ Φ β 1 α 2 α ′ Ψ ,α ′ Φ ′ refine iteratively, add intermediate nodes if necessary α = α 1 ∩ α 2 β = β 1 ; β 2 7 / 9

  19. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ... Ψ ′ ,α α α 1 β 2 Ψ ,α Ψ ′′ Φ β 1 α 2 α ′ Ψ ,α ′ Φ ′ continue inductively until abstract programs converted to concrete programs α = α 1 ∩ α 2 β = β 1 ; β 2 7 / 9

  20. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model α 1 β 2 Ψ ,α Ψ ′′ Φ β 1 α 2 Problem: New unsatisfied diamonds in generation 1 nodes 7 / 9

  21. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ( X ′ ,γ ′ ) ( X ,γ ) γ ′ γ α 1 β 2 Ψ ,α Ψ ′′ Φ β 1 α 2 Repeat Process: Add witnesses (generation 2), refine 7 / 9

  22. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Construction of the Canonical Model ( X ′ ,γ ′ ) ( X ,γ ) γ ′ γ α 1 β 2 Ψ ,α Ψ ′′ Φ β 1 α 2 Repeat Process: Add witnesses (generation 2), refine All diamonds satified in limit (generation ω ) 7 / 9

  23. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Correctness of the Construction α 1 β 2 Ψ ,α Ψ ′′ Φ β 1 α 2 Need to show: ϕ true at node labelled Φ iff ϕ ∈ Φ 8 / 9

  24. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Correctness of the Construction α 1 β 2 Ψ ,α Ψ ′′ Φ β 1 α 2 In particular: If [ α ]¬ ψ ∈ Ψ ′′ and Ψ ′′ − − α → Ψ, then ψ ∉ Ψ 8 / 9

  25. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Correctness of the Construction a b Ψ ,α Ψ ′′ Φ b c In particular: If [( b ; a ) ∩ ( b ; c )]¬ P ∈ Ψ ′′ and Ψ ′′ ( b ; a )∩( b ; c ) − − − − − − − − − − − → Ψ, then P ∉ Ψ 8 / 9

  26. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Correctness of the Construction a b Ψ ,α Φ Ψ ′′ b c In particular: If [( b ; a ) ∩ ( b ; c )]¬ P ∈ Ψ ′′ and Ψ ′′ ( b ; a )∩( b ; c ) − − − − − − − − − − − → Ψ, then P ∉ Ψ Problem: Program unplanned: structure constructed for b ; ( a ∩ c ) 8 / 9

  27. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 Correctness of the Construction a b Ψ ,α Ψ ′′ Φ b c In particular: If [( b ; a ) ∩ ( b ; c )]¬ P ∈ Ψ ′′ and Ψ ′′ ( b ; a )∩( b ; c ) − − − − − − − − − − − → Ψ, then P ∉ Ψ Problem: Program unplanned: structure constructed for b ; ( a ∩ c ) Can rewrite: [( b ; a ) ∩ ( b ; c )]¬ P → [ b ; ( a ∩ c )]¬ P Correctness of construction provable 8 / 9

  28. Florian Bruse, Daniel Kernberger, Martin Lange: A Canonical Model for PDL 0 End of Talk Further work: • Extend to full PDL with intersection, i.e., with Kleene star (weak completeness only) • Compare present work to existing constructions more thoroughly Thanks for listening! 9 / 9

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend