1/30
Refinements and Abstractions of Signal-Event (Timed) Languages
Paul Gastin
LSV ENS de Cachan & CNRS Paul.Gastin@lsv.ens-cachan.fr
Refinements and Abstractions of Signal-Event (Timed) Languages Paul - - PowerPoint PPT Presentation
Refinements and Abstractions of Signal-Event (Timed) Languages Paul Gastin LSV ENS de Cachan & CNRS Paul.Gastin@lsv.ens-cachan.fr Joint work with B eatrice B erard and Antoine Petit FORMATS, Sept. 26th, 2006 1/30 Outline
1/30
LSV ENS de Cachan & CNRS Paul.Gastin@lsv.ens-cachan.fr
2/30
1
3/30
refinement
abstraction
refinement
refinement
refinement
3/30
refinement
abstraction
refinement
refinement
refinement
4/30
refinement
abstraction
abstraction
5/30
refinement
abstraction
5/30
refinement
abstraction
6/30
2
7/30
◮ Σe finite set of (instantaneous) events ◮ Σs finite set of signals ◮ T time domain, T = T ∪ {∞} ◮ Σ = Σe ∪ (Σs × T) ◮ Notation: ad for (a, d) ∈ Σs × T ◮ Σ∞ set of signal-event (timed) words
◮ Signal stuttering: a2a3 ≈ a5,
7/30
◮ Σe finite set of (instantaneous) events ◮ Σs finite set of signals ◮ T time domain, T = T ∪ {∞} ◮ Σ = Σe ∪ (Σs × T) ◮ Notation: ad for (a, d) ∈ Σs × T ◮ Σ∞ set of signal-event (timed) words
◮ Signal stuttering: a2a3 ≈ a5,
7/30
◮ Σe finite set of (instantaneous) events ◮ Σs finite set of signals ◮ T time domain, T = T ∪ {∞} ◮ Σ = Σe ∪ (Σs × T) ◮ Notation: ad for (a, d) ∈ Σs × T ◮ Σ∞ set of signal-event (timed) words
◮ Signal stuttering: a2a3 ≈ a5,
8/30
◮ Useful to hide signals:
hiding signals
◮ τ 0 ≈ ε : an hidden signal with zero duration is not observable.
◮ Signal-event words SE(Σ) = Σ∞/ ≈
8/30
◮ Useful to hide signals:
hiding signals
◮ τ 0 ≈ ε : an hidden signal with zero duration is not observable.
◮ Signal-event words SE(Σ) = Σ∞/ ≈
8/30
◮ Useful to hide signals:
hiding signals
◮ τ 0 ≈ ε : an hidden signal with zero duration is not observable.
◮ Signal-event words SE(Σ) = Σ∞/ ≈
9/30
◮ States emit signals ◮ Transitions emit (instantaneous) events
◮ Run : Idle3 · Req · Wait2 · TimeOut0 · Req · Wait1 · Ack · Connected8 ◮ SEL : languages accepted by SE-automata without ε-transitions. ◮ SELε : languages accepted by SE-automata with ε-transitions.
10/30
3
11/30
◮ Abstract alphabet : Σe and Σs ◮ Concrete alphabet : Σ′
e and Σ′ s
◮ Substitution σ from SE(Σ) to SE(Σ′) defined by:
e ∪ Σ′ s × {0})∗
11/30
◮ Abstract alphabet : Σe and Σs ◮ Concrete alphabet : Σ′
e and Σ′ s
◮ Substitution σ from SE(Σ) to SE(Σ′) defined by:
e ∪ Σ′ s × {0})∗
12/30
12/30
13/30
4
14/30
◮ σ is a SEL-substitution if each La is in SEL ◮ σ is a SELε-substitution if each La is in SELε
15/30
◮ L = {a0f} is recognized by
◮ La = {b} × T is recognized by
◮ Lf = {c0g} is recognized by
◮ σ(L) = {b0c0g} cannot be accepted without ε-transitions.
e((Σ′ s × {0})Σ′ e)∗
15/30
◮ L = {a0f} is recognized by
◮ La = {b} × T is recognized by
◮ Lf = {c0g} is recognized by
◮ σ(L) = {b0c0g} cannot be accepted without ε-transitions.
e((Σ′ s × {0})Σ′ e)∗
16/30
e((Σ′ s × {0})Σ′ e)∗
16/30
e((Σ′ s × {0})Σ′ e)∗
17/30
e((Σ′ s × {0})Σ′ e)∗
17/30
e((Σ′ s × {0})Σ′ e)∗
18/30
e((Σ′ s × {0})Σ′ e)∗
18/30
e((Σ′ s × {0})Σ′ e)∗
19/30
e((Σ′ s × {0})Σ′ e)∗
19/30
e((Σ′ s × {0})Σ′ e)∗
20/30
e((Σ′ s × {0})Σ′ e)∗
20/30
e((Σ′ s × {0})Σ′ e)∗
21/30
22/30
◮ Let σ : A → P(B∗) be a rational substitution ◮ Let ΠA : (A ⊎ B)∗ → A∗ and ΠB : (A ⊎ B)∗ → B∗ be the projections ◮ Let M =
a∈A
◮ Then, σ(L) = ΠB(Π−1
A (L) ∩ M).
A
◮ This proof technique also applies to inverse substitutions:
B (L) ∩ M).
22/30
◮ Let σ : A → P(B∗) be a rational substitution ◮ Let ΠA : (A ⊎ B)∗ → A∗ and ΠB : (A ⊎ B)∗ → B∗ be the projections ◮ Let M =
a∈A
◮ Then, σ(L) = ΠB(Π−1
A (L) ∩ M).
A
◮ This proof technique also applies to inverse substitutions:
B (L) ∩ M).
23/30
◮ Let ˆ
e and ˆ
s.
◮ Let Π1 : SE(ˆ
e,
s × T.
◮ We will show that for a suitable SELε-language M we have
1 (L) ∩ M)
2 (L) ∩ M)
◮ The class SELε is closed under projection, inverse projection and intersection.
24/30
25/30
1 (L).
26/30
◮ for L ⊆ SE(Σ), we have σ(L) = π2(π−1
1 (L) ∩ M),
◮ for L ⊆ SE(Σ′), we have σ−1(L) = π1(π−1
2 (L) ∩ M).
◮ σ(L) ⊆ π2(π−1
1 (L) ∩ M):
1 (L) ∩ M and v ∈ π2(π−1 1 (L) ∩ M).
◮ π2(π−1
1 (L) ∩ M) ⊆ σ(L):
1 (L) ∩ M) and let w ∈ π−1 1 (L) ∩ M with π2(w) = v.
26/30
◮ for L ⊆ SE(Σ), we have σ(L) = π2(π−1
1 (L) ∩ M),
◮ for L ⊆ SE(Σ′), we have σ−1(L) = π1(π−1
2 (L) ∩ M).
◮ σ(L) ⊆ π2(π−1
1 (L) ∩ M):
1 (L) ∩ M and v ∈ π2(π−1 1 (L) ∩ M).
◮ π2(π−1
1 (L) ∩ M) ⊆ σ(L):
1 (L) ∩ M) and let w ∈ π−1 1 (L) ∩ M with π2(w) = v.
27/30
0f1b0 1f2 · · · b0 n ∈ σ(f)} · f
0 f1bd1 1 f2 · · · ∈ σ(ad0+d1+···)}
28/30
◮ Let Σs = Σ′
s = {a, b} and Σe = Σ′ e = {f}.
◮ Let σ be the SEL-substitution defined by
◮ L = {a1fb0} is a SEL. ◮ σ−1(L) = {a1b0} is not a SEL.
28/30
◮ Let Σs = Σ′
s = {a, b} and Σe = Σ′ e = {f}.
◮ Let σ be the SEL-substitution defined by
◮ L = {a1fb0} is a SEL. ◮ σ−1(L) = {a1b0} is not a SEL.
29/30
5
30/30
◮ Signal-event words are the natural objects for studying refinements,
◮ Extending classical results to SE-automata is not always easy due to
◮ We have proved closure properties (refinement, abstraction) for the general