On dynamic topological and metric logics Roman Kontchakov - - PowerPoint PPT Presentation

on dynamic topological and metric logics
SMART_READER_LITE
LIVE PREVIEW

On dynamic topological and metric logics Roman Kontchakov - - PowerPoint PPT Presentation

On dynamic topological and metric logics Roman Kontchakov Department of Computer Science , Kings College London http://www.dcs.kcl.ac.uk/staff/romanvk joint work with Boris Konev, Frank Wolter and Michael Zakharyaschev Dynamic systems .


slide-1
SLIDE 1

On dynamic topological and metric logics

Roman Kontchakov

Department of Computer Science, King’s College London http://www.dcs.kcl.ac.uk/staff/romanvk joint work with

Boris Konev, Frank Wolter and Michael Zakharyaschev

slide-2
SLIDE 2

Dynamic systems

. . ‘space’ + f

x f( x ) f 2( x )

AiML 2004 11.09.04 1

slide-3
SLIDE 3

Dynamic systems

. . ‘space’ + f

x f( x ) f 2( x )

. . 1 2

x y f( x ) f( y ) f 2( x ) f 2( y )

f f

AiML 2004 11.09.04 1

slide-4
SLIDE 4

Dynamic systems

. . ‘space’ + f

x f( x ) f 2( x )

. . 1 2

x y f( x ) f( y ) f 2( x ) f 2( y )

f f Temporal logic to describe and reason about behaviour of dynamic systems:

  • variables p are interpreted by sets of points, i.e., point x is in p:

x ∈ p

  • x always stays in p:

x ∈ ✷Fp

  • x occurs in p infinitely often:

x ∈ ✷F✸Fp

  • . . .

AiML 2004 11.09.04 1

slide-5
SLIDE 5

Dynamic topological logic

Dynamic topological structure F = T, f

T = T, I

a topological space

T is the universe of T I is the interior operator on T C is the closure operator on T ( CX = −I − X )

f : T → T

a total continuous function

( X open ⇒ f −1(X ) open )

AiML 2004 11.09.04 2

slide-6
SLIDE 6

Dynamic topological logic

Dynamic topological structure F = T, f

T = T, I

a topological space

T is the universe of T I is the interior operator on T C is the closure operator on T ( CX = −I − X )

f : T → T

a total continuous function

( X open ⇒ f −1(X ) open )

Dynamic topo-logic DT L

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • modal (topological) operators I and C
  • temporal operators

, ✷F and ✸F AiML 2004 11.09.04 2

slide-7
SLIDE 7

Dynamic topological logic

Dynamic topological structure F = T, f

T = T, I

a topological space

T is the universe of T I is the interior operator on T C is the closure operator on T ( CX = −I − X )

f : T → T

a total continuous function

( X open ⇒ f −1(X ) open )

Dynamic topo-logic DT L

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • modal (topological) operators I and C
  • temporal operators

, ✷F and ✸F

V a valuation: subsets of T −, ∩ and ∪ I and C V(ϕ) = f −1(V(ϕ))

AiML 2004 11.09.04 2

slide-8
SLIDE 8

Dynamic topological logic

Dynamic topological structure F = T, f

T = T, I

a topological space

T is the universe of T I is the interior operator on T C is the closure operator on T ( CX = −I − X )

f : T → T

a total continuous function

( X open ⇒ f −1(X ) open )

Dynamic topo-logic DT L

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • modal (topological) operators I and C
  • temporal operators

, ✷F and ✸F

V a valuation: subsets of T −, ∩ and ∪ I and C V(ϕ) = f −1(V(ϕ)) V(✷Fϕ) =

  • n=1

f −n(V(ϕ)) and V(✸Fϕ) =

  • n=1

f −n(V(ϕ))

AiML 2004 11.09.04 2

slide-9
SLIDE 9

Classes of dynamic topological structures Topological spaces T = T, I

  • arbitrary topologies
  • Aleksandrov:

arbitrary (not only finite) intersections of open sets are open

— every Kripke frame G = U, R, where R is a quasi-order, induces the Aleksandrov topological space U, IG: IGX = {x ∈ U | ∀y (xRy → y ∈ X )} — conversely, every Aleksandrov space is induced by a quasi-order

  • Euclidean spaces Rn, n ≥ 1
  • . . .

AiML 2004 11.09.04 3

slide-10
SLIDE 10

Classes of dynamic topological structures Topological spaces T = T, I

  • arbitrary topologies
  • Aleksandrov:

arbitrary (not only finite) intersections of open sets are open

— every Kripke frame G = U, R, where R is a quasi-order, induces the Aleksandrov topological space U, IG: IGX = {x ∈ U | ∀y (xRy → y ∈ X )} — conversely, every Aleksandrov space is induced by a quasi-order

  • Euclidean spaces Rn, n ≥ 1
  • . . .

Functions f : T → T

  • continuous
  • homeomorphisms: continuous bijections with continuous inverses
  • . . .

AiML 2004 11.09.04 3

slide-11
SLIDE 11

Known results

DT L — subset of DT L containing no ‘infinite’ operators (✷F and ✸F) Artemov, Davoren & Nerode (1997): The two dynamic topo-logics Log{F, f} and Log{F, f | F an Aleksandrov space} coincide, have the fmp, are finitely axiomatisable, and so decidable. NB. Log{F, f} Log{R, f} (Slavnov 2003, Kremer & Mints 2003)

AiML 2004 11.09.04 4

slide-12
SLIDE 12

Known results

DT L — subset of DT L containing no ‘infinite’ operators (✷F and ✸F) Artemov, Davoren & Nerode (1997): The two dynamic topo-logics Log{F, f} and Log{F, f | F an Aleksandrov space} coincide, have the fmp, are finitely axiomatisable, and so decidable. NB. Log{F, f} Log{R, f} (Slavnov 2003, Kremer & Mints 2003) Kremer, Mints & Rybakov (1997): The three dynamic topo-logics Log{F, f | f a homeomorphism}, Log{F, f | F an Aleksandrov space, f a homeomorphism}, Log{R, f | f a homeomorphism} coincide, have the fmp, are finitely axiomatisable, and so decidable.

AiML 2004 11.09.04 4

slide-13
SLIDE 13

Known results

DT L — subset of DT L containing no ‘infinite’ operators (✷F and ✸F) Artemov, Davoren & Nerode (1997): The two dynamic topo-logics Log{F, f} and Log{F, f | F an Aleksandrov space} coincide, have the fmp, are finitely axiomatisable, and so decidable. NB. Log{F, f} Log{R, f} (Slavnov 2003, Kremer & Mints 2003) Kremer, Mints & Rybakov (1997): The three dynamic topo-logics Log{F, f | f a homeomorphism}, Log{F, f | F an Aleksandrov space, f a homeomorphism}, Log{R, f | f a homeomorphism} coincide, have the fmp, are finitely axiomatisable, and so decidable. Open problem: axiomatisations and algorithmic properties of the full DT L ?

AiML 2004 11.09.04 4

slide-14
SLIDE 14

Homeomorphisms: bad news

Theorem 1. No logic from the list below is recursively enumerable:

  • Log {F, f | f a homeomorphism},
  • Log {F, f | F an Aleksandrov space, f a homeomorphism},
  • Log {Rn, f | f a homeomorphism, n ≥ 1}.
  • Proof. By reduction of the undecidable but r.e. Post’s Correspondence Problem

to the satisfiability problem.

AiML 2004 11.09.04 5

slide-15
SLIDE 15

Homeomorphisms: bad news

Theorem 1. No logic from the list below is recursively enumerable:

  • Log {F, f | f a homeomorphism},
  • Log {F, f | F an Aleksandrov space, f a homeomorphism},
  • Log {Rn, f | f a homeomorphism, n ≥ 1}.
  • Proof. By reduction of the undecidable but r.e. Post’s Correspondence Problem

to the satisfiability problem. NB. All these logics are different.

AiML 2004 11.09.04 5

slide-16
SLIDE 16

Continuous maps: some good news

Finite iterations:

  • arbitrary finite flows of time
  • finite change assumption

(the system eventually stabilises)

AiML 2004 11.09.04 6

slide-17
SLIDE 17

Continuous maps: some good news

Finite iterations:

  • arbitrary finite flows of time
  • finite change assumption

(the system eventually stabilises)

Theorem 2. The two topo-logics Log∗ {F, f} and Log∗ {F, f | F an Aleksandrov space} coincide and are decidable, but not in primitive recursive time.

  • Proof. By Kruskal’s tree theorem

and reduction of the reachability problem for lossy channels.

AiML 2004 11.09.04 6

slide-18
SLIDE 18

Continuous maps: some good news

Finite iterations:

  • arbitrary finite flows of time
  • finite change assumption

(the system eventually stabilises)

Theorem 2. The two topo-logics Log∗ {F, f} and Log∗ {F, f | F an Aleksandrov space} coincide and are decidable, but not in primitive recursive time.

  • Proof. By Kruskal’s tree theorem

and reduction of the reachability problem for lossy channels. However: Theorem 3. The two topo-logics Log∗ {F, f | f a homeomorphism} and Log∗ {F, f | F an Aleksandrov space, f a homeomorphism} coincide but are not recursively enumerable.

AiML 2004 11.09.04 6

slide-19
SLIDE 19

Dynamics in metric spaces

. . 1 2

x y f( x ) f( y ) f 2( x ) f 2( y )

f f

AiML 2004 11.09.04 7

slide-20
SLIDE 20

Dynamics in metric spaces

. . 1 2

x y f( x ) f( y ) f 2( x ) f 2( y )

f f A metric space D = W, d, where d: W × W → R+ is a metric, induces the topological space Td = W, Id: IdX = {x ∈ W | ∃δ > 0 ∀y ∈ W (d(x, y) < δ → y ∈ X)}

AiML 2004 11.09.04 7

slide-21
SLIDE 21

Logics of metric spaces

D = W, d

a metric space

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • interior I and closure C operators
  • universal ∀ and existential ∃ modalities
  • metric operators ∃≤a and ∀≤a,

for a ∈ Q+

AiML 2004 11.09.04 8

slide-22
SLIDE 22

Logics of metric spaces

D = W, d

a metric space

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • interior I and closure C operators
  • universal ∀ and existential ∃ modalities
  • metric operators ∃≤a and ∀≤a,

for a ∈ Q+ V a valuation: subsets of W −, ∩ and ∪ Id and Cd . . V(∃≤aϕ) = {x ∈ W | ∃y ∈ V(ϕ) such that d(x, y) ≤ a} V(∀≤aϕ) = {x ∈ W | ∀y ∈ V(ϕ) such that d(x, y) ≤ a} V(ϕ)

a a AiML 2004 11.09.04 8

slide-23
SLIDE 23

Logics of metric spaces

D = W, d

a metric space

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • interior I and closure C operators
  • universal ∀ and existential ∃ modalities
  • metric operators ∃≤a and ∀≤a,

for a ∈ Q+ V a valuation: subsets of W −, ∩ and ∪ Id and Cd . . V(∃≤aϕ) = {x ∈ W | ∃y ∈ V(ϕ) such that d(x, y) ≤ a} V(∀≤aϕ) = {x ∈ W | ∀y ∈ V(ϕ) such that d(x, y) ≤ a} V(ϕ)

a a

Wolter & Zakharyaschev (2004): The set of valid MT -formulas is axiomatisable. The satisfiability of MT -formulas is EXPTIME-complete.

AiML 2004 11.09.04 8

slide-24
SLIDE 24

Dynamic metric logic

Dynamic metric structure F = D, f

D = W, d

a metric space

f : W → W

a metric automorphism

(bijection, d(f(x), f(y)) = d(x, y))

AiML 2004 11.09.04 9

slide-25
SLIDE 25

Dynamic metric logic

Dynamic metric structure F = D, f

D = W, d

a metric space

f : W → W

a metric automorphism

(bijection, d(f(x), f(y)) = d(x, y))

Dynamic metric logic DML

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • modal (metric) operators ∃≤a and ∀≤a,

for a ∈ Q+

  • interior I and closure C operators
  • universal ∀ and existential ∃ modalities
  • temporal operators

, ✷F and ✸F AiML 2004 11.09.04 9

slide-26
SLIDE 26

Dynamic metric logic

Dynamic metric structure F = D, f

D = W, d

a metric space

f : W → W

a metric automorphism

(bijection, d(f(x), f(y)) = d(x, y))

Dynamic metric logic DML

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • modal (metric) operators ∃≤a and ∀≤a,

for a ∈ Q+

  • interior I and closure C operators
  • universal ∀ and existential ∃ modalities
  • temporal operators

, ✷F and ✸F AiML 2004 11.09.04 9

slide-27
SLIDE 27

Dynamic metric logic

Dynamic metric structure F = D, f

D = W, d

a metric space

f : W → W

a metric automorphism

(bijection, d(f(x), f(y)) = d(x, y))

Dynamic metric logic DML

  • propositional variables p, q, . . .
  • the Booleans ¬, ∧ and ∨
  • modal (metric) operators ∃≤a and ∀≤a,

for a ∈ Q+

  • interior I and closure C operators
  • universal ∀ and existential ∃ modalities
  • temporal operators

, ✷F and ✸F

Theorem 4. The set of valid DML-formulas is decidable. However, the decision problem is not elementary. Proof. Quasimodels, reduction to monadic second-order logic and yardsticks.

AiML 2004 11.09.04 9

slide-28
SLIDE 28

Open problems and future research The field still remains a big research challenge. . .

  • Axiomatisation of decidable logics/fragments
  • Various topological and metric spaces: Euclidean, compact, etc.
  • functions: Lipschitz continuous, contracting maps, etc.
  • Model checking
  • . . .

AiML 2004 11.09.04 10