Expressiveness, decidability, and undecidability of ...users.dimi.uniud.it/.../thesis/phd_thesis_slides.pdf · D. Della Monica University of Udine Department of Mathematics and Computer

D. Della Monica

University of Udine

Department of Mathematics and Computer Science

Expressiveness, decidability, and

undecidability of Interval Temporal Logic

ITL - Beyond the end of the light

Ph.D. Defence

Dario Della Monica

supervisor: A. Montanarico-supervisors: G. Sciavicco and V. Goranko

Udine - April 1, 2011

D. Della Monica

At the beginning...

At the beginning, it was the darkness...

Then, logicians made the light,

they became curious,

and moved toward the darkness...

... as close as they could

D. Della Monica

Outline

IntroductionThe Halpern and Shoham’s logic HS

Expressiveness of HS

The satisfiability problem for HSUndecidability

Classical extensionsMetric extensionsHybrid extensionsFirst-order extensions

Summary and perspectives

D. Della Monica

Outline






D. Della Monica

Interval-based temporal reasoning: origins and applications

Interval-based temporal reasoning: reasoning about time, where theprimary concept is ‘time interval’, rather than ‘time instant’.

Origins:

◮ Philosophy, in particular philosophy and ontology of time.

◮ Linguistics: analysis of progressive tenses, semantics of naturallanguages.

◮ Artificial intelligence: temporal knowledge representation,temporal planning, theory of events, etc.

◮ Computer science: specification and design of hardwarecomponents, concurrent real-time processes, temporaldatabases, etc.

D. Della Monica

Motivations

Some properties are intrinsically related to a time interval instead ofa (set of) time instant, each of which has not duration.

D. Della Monica

Motivations


Think about the event: “traveling from A to B”:

D. Della Monica

Motivations



◮ it is true over a precise interval of time

◮ it is not true over any other interval (starting/ending interval,inner interval, ecc.)

D. Della Monica

Motivations





Several philosophical and logical paradoxes disappear:

D. Della Monica

Motivations






◮ Zeno’s flying arrow paradox (“if at each instant the flyingarrow stands still, how is movement possible?”)

◮ The dividing instant dilemma (“if the light is on and it is turnedoff, what is its state at the instant between the two events?”)

D. Della Monica

Motivations






◮ Zeno’s flying arrow paradox (“if at each instant the flyingarrow stands still, how is movement possible?”)

◮ The dividing instant dilemma (“if the light is on and it is turnedoff, what is its state at the instant between the two events?”)

The truth of a formula over an interval does not necessarily dependon its truth over subintervals.

D. Della Monica

Interval temporal reasoning and temporal ontologies

Interval-based temporal reasoning is subject to the same ontologicaldilemmas regarding the nature of Time as the instant-basedtemporal reasoning:

D. Della Monica



◮ linear or branching?

D. Della Monica




◮ discrete or dense?

D. Della Monica





◮ with or without beginning/end?, etc.

D. Della Monica






New issues arise regarding the nature of the intervals:

D. Della Monica







◮ Can intervals be unbounded?

D. Della Monica







◮ Can intervals be unbounded?

◮ Are intervals with coinciding endpoints admissible or not?

D. Della Monica

Intervals and interval structures

D = 〈D, <〉: partially ordered set.

An interval in D: ordered pair [a, b], where a, b ∈ D and a ≤ b.

D. Della Monica




If a < b then [a, b] is a strict interval; [a, a] is a point interval.

D. Della Monica





I(D): the interval structure over D, consisting of the set of allintervals over D.

D. Della Monica






In this talk I will restrict attention to linear interval structures,i.e. interval structures over linear orders.

D. Della Monica






In this talk I will restrict attention to linear interval structures,i.e. interval structures over linear orders.

In particular, standard interval structures on N,Z,Q, and R withtheir usual orders.

D. Della Monica

Binary interval relations on linear orders

J. F. Allen

Maintaining knowledge about temporal intervals.

Communications of the ACM, volume 26(11), pages 832-843, 1983.

D. Della Monica


Later

J. F. Allen



D. Della Monica


Later

After (right neighbour)

J. F. Allen



D. Della Monica


Later


Overlaps (to right)

J. F. Allen



D. Della Monica


Later


Overlaps (to right)

Ends

J. F. Allen



D. Della Monica


Later


Overlaps (to right)

Ends

During (subinterval)

J. F. Allen



D. Della Monica


Later


Overlaps (to right)

Ends


Begins

J. F. Allen



D. Della Monica


Later


Overlaps (to right)

Ends


Begins

6 relations + their inverses + equality = 13 Allen’s relations.

J. F. Allen



D. Della Monica

Outline






D. Della Monica

Halpern-Shoham’s modal logic of interval relationsEvery interval relation gives rise to a modal operator over relationalinterval structures.

D. Della Monica

Halpern-Shoham’s modal logic of interval relationsEvery interval relation gives rise to a modal operator over relationalinterval structures. Thus, a multimodal logic arises:

J. Halpern and Y. Shoham

A propositional modal logic of time intervals.

Journal of the ACM, volume 38(4), pages 935-962, 1991.

D. Della Monica





non-strict semantics:All modalities are definable in terms

of 〈B〉,〈E〉,〈B〉,〈E〉

HS ≡ BEBE

D. Della Monica






of 〈B〉,〈E〉,〈B〉,〈E〉

HS ≡ BEBE

〈B〉

〈B〉〈E〉

〈E〉

D. Della Monica






of 〈B〉,〈E〉,〈B〉,〈E〉

HS ≡ BEBE

strict semantics:Also needed additional modalities

〈A〉,〈A〉

HS ≡ BEBE + AA

〈A〉

〈A〉

D. Della Monica






of 〈B〉,〈E〉,〈B〉,〈E〉

HS ≡ BEBE

strict semantics:Also needed additional modalities

〈A〉,〈A〉

HS ≡ BEBE + AA

〈A〉

〈A〉

Syntax of Halpern-Shoham’s logic, hereafter called HS :

φ ::= p | ¬φ | φ ∧ ψ | 〈B〉φ | 〈E〉φ | 〈B〉φ | 〈E〉φ(

| 〈A〉φ | 〈A〉φ)

.

D. Della Monica

Models for propositional interval logics

AP : a set of atomic propositions (over intervals).

D. Della Monica



Non-strict interval model:

M+ = 〈I(D)+,V 〉,

where V : AP 7→ 2I(D)+.

D. Della Monica



Non-strict interval model:

M+ = 〈I(D)+,V 〉,

where V : AP 7→ 2I(D)+.

Strict interval model:

M− = 〈I(D)−,V 〉,

where V : AP 7→ 2I(D)−.

D. Della Monica

Formal semantics of HS

〈B〉: M, [d0, d1] 〈B〉φ iff there exists d2 such that d0 ≤ d2 < d1 andM, [d0, d2] φ.

〈B〉: M, [d0, d1] 〈B〉φ iff there exists d2 such that d1 < d2 andM, [d0, d2] φ.

current interval:

〈B〉φ:φ

〈B〉φ:φ

D. Della Monica




〈E〉: M, [d0, d1] 〈E〉φ iff there exists d2 such that d0 < d2 ≤ d1 andM, [d2, d1] φ.

〈E〉: M, [d0, d1] 〈E〉φ iff there exists d2 such that d2 < d0 andM, [d2, d1] φ.

current interval:

〈E〉φ:φ

〈E〉φ:φ

D. Della Monica




〈E〉: M, [d0, d1] 〈E〉φ iff there exists d2 such that d0 < d2 ≤ d1 andM, [d2, d1] φ.

〈E〉: M, [d0, d1] 〈E〉φ iff there exists d2 such that d2 < d0 andM, [d2, d1] φ.

〈A〉: M, [d0, d1] 〈A〉φ iff there exists d2 such that d1 < d2 andM, [d1, d2] φ.

〈A〉: M, [d0, d1] 〈A〉φ iff there exists d2 such that d2 < d0 andM, [d2, d0] φ.

current interval:

〈A〉φ:φ

〈A〉φ:φ

D. Della Monica

Formal semantics of HS - contd’

〈L〉: M, [d0, d1] 〈L〉φ iff there exists d2, d3 such that d1 < d2 < d3 andM, [d2, d3] φ.


current interval:

〈L〉φ:φ

〈L〉φ:φ

D. Della Monica




〈D〉: M, [d0, d1] 〈D〉φ iff there exists d2, d3 such that d0 < d2 < d3 < d1 andM, [d2, d3] φ.


current interval:

〈D〉φ:φ

〈D〉φ:φ

D. Della Monica






〈O〉: M, [d0, d1] 〈O〉φ iff there exists d2, d3 such that d0 < d2 < d1 < d3 andM, [d2, d3] φ.

〈O〉: M, [d0, d1] 〈O〉φ iff there exists d2, d3 such that d2 < d0 < d3 < d1 andM, [d2, d3] φ.

current interval:

〈O〉φ:φ

〈O〉φ:φ

D. Della Monica

Defining the other interval modalities in HS

A useful new symbol is the modal constant π for point-intervals:

M, [d0, d1] π iff d0 = d1.

D. Della Monica



M, [d0, d1] π iff d0 = d1.

It is definable as either [B ]⊥ or [E ]⊥, so it is only needed in weakerfragments of HS.

D. Della Monica



M, [d0, d1] π iff d0 = d1.

It is definable as either [B ]⊥ or [E ]⊥, so it is only needed in weakerfragments of HS.

In general, it is possible defining HS modalities in terms of others

D. Della Monica

The zoo of fragments of HS

Technically, there are 212 = 4096 fragments of HSOf them, several hundreds are of essentially different expressiveness

D. Della Monica

The zoo of fragments of HS

Technically, there are 212 = 4096 fragments of HSOf them, several hundreds are of essentially different expressiveness

Each of these, considered with respect to some parameters:

1. over special classes of interval structures (all, dense, discrete, finite, etc.)

2. with strict or non-strict semantics

3. including or excluding π operator (whenever it cannot be defined)

D. Della Monica

Outline






D. Della Monica

Comparing the expressiveness of fragments of HS

Expressiveness classification problem: classify the fragments of HSwith respect to their expressiveness, relative to important classes ofinterval models.

D. Della Monica

The problem of comparing expressive power of HS fragmentsL1, L2 HS-fragments

L1 L2

D. Della Monica


L1 {≺,≡,≻, 6≈} L2

D. Della Monica


L1 {≺,≡,≻, 6≈} L2

does L1 translateinto L2?



≡ ≺ ≻ 6≈

yes no

yes no yes no

D. Della Monica

Truth-preserving translation

There exists a truth-preserving translation of L1 into L2

iffL2 is at least as expressive as L1

(L1 � L2)

D. Della Monica




(L1 � L2)

For each modal operator 〈X 〉 of L1 there exists a L2-formula ϕ s.t.

〈X 〉p ≡ ϕ

D. Della Monica




(L1 � L2)

For each modal operator 〈X 〉 of L1 there exists a L2-formula ϕ s.t.

〈X 〉p ≡ ϕ

212 fragments... 212·(212−1)2 comparisons

D. Della Monica

Inter-definability equations

Notation: X1X2 . . .Xn will denote the fragment of

HS containing the modalities 〈X1〉, 〈X2〉, . . . , 〈Xn〉

〈L〉p ≡ 〈A〉〈A〉p〈O〉p ≡ 〈E 〉〈B〉p〈D〉p ≡ 〈E 〉〈B〉p

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB 〈D〉 ⊑ EB

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB 〈D〉 ⊑ EB

〈L〉p ≡ 〈B〉[E ]〈B〉〈E 〉p 〈L〉 ⊑ BE 〈L〉 ⊑ BE

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB 〈D〉 ⊑ EB

〈L〉p ≡ 〈B〉[E ]〈B〉〈E 〉p 〈L〉 ⊑ BE 〈L〉 ⊑ BE

Soundness and completeness???

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB 〈D〉 ⊑ EB

〈L〉p ≡ 〈B〉[E ]〈B〉〈E 〉p 〈L〉 ⊑ BE 〈L〉 ⊑ BE


Soundness: all equations are valid

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB 〈D〉 ⊑ EB

〈L〉p ≡ 〈B〉[E ]〈B〉〈E 〉p 〈L〉 ⊑ BE 〈L〉 ⊑ BE


Soundness: all equations are valid SIMPLE

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB 〈D〉 ⊑ EB

〈L〉p ≡ 〈B〉[E ]〈B〉〈E 〉p 〈L〉 ⊑ BE 〈L〉 ⊑ BE



Completeness:there are no more

inter-definability equations

D. Della Monica




〈L〉p ≡ 〈A〉〈A〉p 〈L〉 ⊑ A 〈L〉 ⊑ A

〈O〉p ≡ 〈E 〉〈B〉p 〈O〉 ⊑ EB 〈O〉 ⊑ EB

〈D〉p ≡ 〈E 〉〈B〉p 〈D〉 ⊑ EB 〈D〉 ⊑ EB

〈L〉p ≡ 〈B〉[E ]〈B〉〈E 〉p 〈L〉 ⊑ BE 〈L〉 ⊑ BE



Completeness:there are no more

BISIMULATIONSinter-definability equations

D. Della Monica

Bisimulation between interval structuresZ ⊆ M1 × M2 is a bisimulations wrt the fragment X1X2 . . .Xn iff

D. Della Monica


1. Z -related intervals satisfy the same propositional letters, i.e.:

(i1, i2) ∈ Z ⇒ (p is true over i1 ⇔ p is true over i2)

D. Della Monica




2. the bisimulation relation is “preserved” by modal operators,i.e., for every modal operator 〈X 〉:

M1

M2

i1

i2

i ′1

D. Della Monica





(i1, i2) ∈ Z

M1

M2

i1

i2

i ′1

Z

D. Della Monica





(i1, i2) ∈ Z

(i1, i′1) ∈ X

M1

M2

i1

i2

i ′1

Z

X

D. Della Monica





(i1, i2) ∈ Z

(i1, i′1) ∈ X

}

⇒ ∃i ′2 s.t.

M1

M2

i1

i2

i ′1

i ′2

Z

X

D. Della Monica





(i1, i2) ∈ Z

(i1, i′1) ∈ X

}

⇒ ∃i ′2 s.t.

{

(i ′1, i′2) ∈ Z

(i2, i′2) ∈ X

M1

M2

i1

i2

i ′1

i ′2

Z

X

Z

X

D. Della Monica

Bisimulation between interval structures - cont’d

Theorem Let Z be a bisimulation between M1 and M2 for thelanguage L and let i1 and i2 be intervals in M1 and M2,respectively. Then, truth of L-formulae is preserved by Z , i.e.,

If (i1, i2) ∈ Z , then for every formula ϕ of L:

M1, i1 ϕ iff M2, i2 ϕ

D. Della Monica

How to use bisimulations to disprove definability

Suppose that we want to prove:

〈X 〉 is not definable in terms of L

D. Della Monica




We must provide:

1. two models M1 and M2

D. Della Monica




We must provide:


2. a bisimulation Z ⊆ M1 × M2 wrt fragment L

D. Della Monica




We must provide:



3. two interval i1 ∈ M1 and i2 ∈ M2 such that

a. i1 and i2 are Z -relatedb. M1, i1 〈X 〉p and M2, i2 ¬〈X 〉p

D. Della Monica




We must provide:





By contradiction

If 〈X 〉 is definable in terms of L, then 〈X 〉p is.

D. Della Monica




We must provide:





By contradiction

If 〈X 〉 is definable in terms of L, then 〈X 〉p is.Truth of 〈X 〉p should have been preserved by Z , but 〈X 〉p is true ini1 (in M1) and false in i2 (in M2)

D. Della Monica




We must provide:





By contradiction

If 〈X 〉 is definable in terms of L, then 〈X 〉p is.Truth of 〈X 〉p should have been preserved by Z , but 〈X 〉p is true ini1 (in M1) and false in i2 (in M2) ⇒ contradiction

D. Della Monica

An example: the operator 〈D〉

Semantics:

M, [a, b] 〈D〉ϕdef⇔ ∃c , d such that a < c < d < b and M, [c , d ] ϕ

〈D〉ϕ

ϕ

D. Della Monica


Semantics:


〈D〉ϕ

ϕ

Operator 〈D〉 is definable in terms of BE 〈D〉ϕ ≡ 〈B〉〈E 〉ϕ

D. Della Monica


Semantics:


〈D〉ϕ

ϕ

Operator 〈D〉 is definable in terms of BE 〈D〉ϕ ≡ 〈B〉〈E 〉ϕ

To prove that 〈D〉 is not definable in terms of any other fragment,we must prove that:

1) 〈D〉 is not definable in terms of ALBOALBEDO

2) 〈D〉 is not definable in terms of ALEOALBEDO

D. Della Monica

〈D〉 is not definable in terms of AA bisimulation wrt fragment A but not D

Bisimulation wrt A (AP = {p}):◮ models: M1 = 〈I(N),V1〉,M2 = 〈I(N),V2〉

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

D. Della Monica



◮ V1(p) = {[1, 2]}

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

p

D. Della Monica



◮ V1(p) = {[1, 2]}◮ V2(p) = ∅

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

p

¬p

D. Della Monica



◮ V1(p) = {[1, 2]}◮ V2(p) = ∅

◮ bisimulation relation Z : ([x , y ], [w , z]) ∈ Z iff

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

p

¬p

D. Della Monica



◮ V1(p) = {[1, 2]}◮ V2(p) = ∅


1. [x , y ] = [w , z] = [0, 3]

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

p

¬p

D. Della Monica



◮ V1(p) = {[1, 2]}◮ V2(p) = ∅


1. [x , y ] = [w , z] = [0, 3]2. [x , y ] = [w , z] and x ≥ 3

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

p

¬p

D. Della Monica



◮ V1(p) = {[1, 2]}◮ V2(p) = ∅


1. [x , y ] = [w , z] = [0, 3]2. [x , y ] = [w , z] and x ≥ 3

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

p

¬p

M1, [0, 3] 〈D〉p and M2, [0, 3] ¬〈D〉p

D. Della Monica



◮ V1(p) = {[1, 2]}◮ V2(p) = ∅


1. [x , y ] = [w , z] = [0, 3]2. [x , y ] = [w , z] and x ≥ 3

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9

p

¬p

M1, [0, 3] 〈D〉p and M2, [0, 3] ¬〈D〉p ⇒ the thesis

D. Della Monica

Outline






D. Della Monica

The satisfiability problem for HS

Satisfiability problem for a logic L Given an L-formula ϕ, is ϕsatisfiable, i.e., there exists a model and an interval in which ϕ istrue?

D. Della Monica



L is decidable (wrt the satisfiability problem)iff

for each formula it is possible to answer the question

D. Della Monica




for each formula it is possible to answer the questionthere exists a terminating algorithm that answer yes / not for any ϕ

D. Della Monica





Expressive enough, yet decidable, HS fragments

D. Della Monica





Expressive enough, yet decidable, HS fragments

Classification of all HS fragments wrt (un)decidability

D. Della Monica

Maximal decidable HS fragments

◮ PNL (≡ AA) in general case

D. Bresolin, V. Goranko, A. Montanari, G. Sciavicco

Propositional Interval Neighborhood Logic: Decidability,Expressiveness, and Undecidable Extensions.

Annals of Pure and Applied Logics, 2009, 161, 289-304.

D. Della Monica



◮ ABBL (and AEEL) in general case

D. Bresolin, A. Montanari, P. Sala, G. Sciavicco

What’s decidable about Halpern and Shoham’s interval logic?The maximal fragment ABBL.

LICS 2011, 2011.

D. Della Monica




◮ ABBA (and AEEA) over finite structures

A. Montanari, G. Puppis, P. Sala

Maximal Decidable Fragments of Halpern and Shoham’sModal Logic of Intervals.

ICALP 2010, 2010, LNCS 6199, 2010, 345-356.

D. Della Monica




◮ ABBA (and AEEA) over finite structures

◮ DDBBLL over Q

P. Sala

PhD thesis

2010

D. Della Monica

Weakest undecidable HS fragments

AD,AD,AD,AD

D. Della Monica


AD,AD,AD,AD

BE,BE,BE,BE

D. Della Monica


AD,AD,AD,AD

BE,BE,BE,BE

O (and O)

D. Della Monica


AD,AD,AD,AD

BE,BE,BE,BE

O (and O)

D (and D) over discrete

D. Della Monica


AD,AD,AD,AD [in this thesis]

BE,BE,BE,BE [in this thesis]

O (and O) [in this thesis]

D (and D) over discrete [Michaliszyn, Marcinkowski]

The Ultimate Undecidability Result

for the Halpern-Shoham Logic

LICS 2011

D. Della Monica

Outline






D. Della Monica

The Octant Tiling ProblemThis is the problem of establishing whether a given finite set of tiletypes T = {t1, . . . , tk} can tile the 2nd octant of the integer plane:

O = {(i , j) : i , j ∈ N ∧ 0 ≤ i ≤ j},

while respecting the color constraints.

D. Della Monica


O = {(i , j) : i , j ∈ N ∧ 0 ≤ i ≤ j},


D. Della Monica


O = {(i , j) : i , j ∈ N ∧ 0 ≤ i ≤ j},


Proposition The Octant Tiling Problem is undecidable.

D. Della Monica


O = {(i , j) : i , j ∈ N ∧ 0 ≤ i ≤ j},


Proposition The Octant Tiling Problem is undecidable.

Proof: by reduction from the tiling problem for N× N, usingKönig’s Lemma.

D. Della Monica

Undecidability of the interval logics via tiling:generic construction

1. Encoding of the octant

2. Encoding of the neighborhood relations◮ Right-neighborhood relation SIMPLE◮ Above-neighborhood relation HARD

D. Della Monica




Encoding of the octant

◮ Force the existence of a unique infinite chain of unit-intervalson the linear order, which covers an initial segment of theinterval model. (propositional letter u)

Unit intervals are used to place tiles and delimiting symbols.

D. Della Monica




Encoding of the octant

◮ Force the existence of a unique infinite chain of unit-intervalson the linear order, which covers an initial segment of theinterval model. (propositional letter u)

Unit intervals are used to place tiles and delimiting symbols.

◮ ID-intervals are then introduced to represent the layers of tiles.(propositional letter Id)

D. Della Monica

Undecidability of the interval logics via tiling:generic construction cont’d

D. Della Monica


Each ID-interval must have the right number of tiles

D. Della Monica



The most challenging part usually is to ensure that the consecutiveID-intervals match vertically: the Above-Neighbour relation.

D. Della Monica



The most challenging part usually is to ensure that the consecutiveID-intervals match vertically: the Above-Neighbour relation.

For that, auxiliary propositional letter up_rel can be used toconnecting (endpoints of) two intervals representing tiles that areabove connected in the octant

D. Della Monica

Undecidability of the interval logics via tiling:generic construction completed

Eventually, we encode the given Octant tiling problem by specifyingthe matching conditions between intervals that are right-connectedor above-connected.

D. Della Monica

Undecidability of the interval logics via tiling:generic construction completed

Eventually, we encode the given Octant tiling problem by specifyingthe matching conditions between intervals that are right-connectedor above-connected.

The specific part of the construction is to use the given fragment ofHS to set the chain of unit intervals and to express all necessaryproperties of IDs, the propositional letters for correspondenceintervals, and the tile matching conditions.

D. Della Monica

Summary of (un)decidability results and outlook

◮ In summary: interval logics are generally undecidable, evenunder very weak assumptions.

D. Della Monica



◮ In particular, most fragments of HS (∼ 90%) have beenproved undecidable over most of the natural classes of intervalstructures.

D. Della Monica




◮ There are still some currently unknown cases (< 8%), andsome are conjectured undecidable (over general and specialclasses), e.g., L∗D∗, where X ∗ ∈ {X ,X}; etc.

D. Della Monica





◮ Not all results transfer readily between the strict and thenon-strict semantics, and between the classes of all, dense,discrete, etc. interval structures.

D. Della Monica





◮ Not all results transfer readily between the strict and thenon-strict semantics, and between the classes of all, dense,discrete, etc. interval structures.

◮ More statistics are available on the web page:https://itl.dimi.uniud.it/content/logic-hs

D. Della Monica

Outline






D. Della Monica

Outline






D. Della Monica

PNL: syntax and semantics

Syntax

◮ PNL: ϕ ::= p | ¬ϕ | ϕ ∨ ϕ | 〈A〉ϕ | 〈A〉ϕ

D. Della Monica


Syntax

◮ PNL: ϕ ::= p | ¬ϕ | ϕ ∨ ϕ | 〈A〉ϕ | 〈A〉ϕ

Semantics

◮ Operators meets (〈A〉) and met-by (〈A〉):

meets:〈A〉ϕ

D. Della Monica


Syntax

◮ PNL: ϕ ::= p | ¬ϕ | ϕ ∨ ϕ | 〈A〉ϕ | 〈A〉ϕ

Semantics


meets:〈A〉ϕ ϕ

D. Della Monica


Syntax

◮ PNL: ϕ ::= p | ¬ϕ | ϕ ∨ ϕ | 〈A〉ϕ | 〈A〉ϕ

Semantics


meets:〈A〉ϕ ϕ

met-by:〈A〉ϕ

D. Della Monica


Syntax

◮ PNL: ϕ ::= p | ¬ϕ | ϕ ∨ ϕ | 〈A〉ϕ | 〈A〉ϕ

Semantics


meets:〈A〉ϕ ϕ

met-by:〈A〉ϕϕ

D. Della Monica

Metric interval logics

Two types of metric extensions of interval logics over the integers:

D. Della Monica



1. Extensions of the modal operators: 〈A〉=k , 〈A〉>k , 〈A〉[k,k′], . . .

D. Della Monica




2. Atomic propositions for length constraints: len>k, len=k, . . .

D. Della Monica





The former are definable in terms of the latter in PNL, e.g.:

〈A〉>kp := 〈A〉(p ∧ len>k).

D. Della Monica





The former are definable in terms of the latter in PNL, e.g.:

〈A〉>kp := 〈A〉(p ∧ len>k).

MPNL: PNL extended with integer constraints for interval lengths.

D. Della Monica

Decidability of metric interval logic

Theorem Satisfiability in MPNL on N is decidable. It isNEXPTIME-complete if the metric constraints are represented inunary, and in between EXPSPACE and 2NEXPTIME if they arerepresented in binary.

D. Bresolin, D. Della Monica, V. Goranko, A. Montanari, G. Sciavicco

Metric Propositional Neighborhood Interval Logics: expressiveness,decidability, and undecidability, Proc. of the European Conference onArtificial Intelligence (ECAI), 2010.

D. Della Monica

Decidability of metric interval logic

Theorem Satisfiability in MPNL on N is decidable. It isNEXPTIME-complete if the metric constraints are represented inunary, and in between EXPSPACE and 2NEXPTIME if they arerepresented in binary.

D. Bresolin, D. Della Monica, V. Goranko, A. Montanari, G. Sciavicco

Metric Propositional Neighborhood Interval Logics: expressiveness,decidability, and undecidability, Proc. of the European Conference onArtificial Intelligence (ECAI), 2010.

Exact complexity is an open problem

D. Della Monica

Relative expressive power of logics in MPNL

MPNL<

MPNL=

MPNL>

MPNL()

MPNL=l MPNL

()

l

D. Della Monica


MPNLε ≡ PNL

MPNL<,ε MPNL>,ε

MPNL=,ε MPNL<,> MPNL(),ε

MPNL[]

MPNL

MPNL<

MPNL=

MPNL>

MPNL()

MPNL=l MPNL

()

l

D. Della Monica


MPNLε ≡ PNL

MPNL<,ε MPNL>,ε


MPNL[]

MPNL

MPNL<

MPNL=

MPNL>

MPNL()

MPNL=l MPNL

()

l

in 2NEXPTIME

EXPSPACE -hard

NEXPTIME -complete

D. Della Monica


MPNLε ≡ PNL

MPNL<,ε MPNL>,ε


MPNL[]

MPNL

MPNL<

MPNL=

MPNL>

MPNL()

MPNL=l MPNL

()

l

in 2NEXPTIME

EXPSPACE -hard

NEXPTIME -complete

Decidability of MPNL: by small model propertyComparing expressiveness of metric fragments: by bisimulations

D. Della Monica

Outline






D. Della Monica

Extending PNL

PNL

D. Della Monica

Extending PNL

PNL

NEXPTIME-co

D. Della Monica

Extending PNL

PNL

PNL +any HS operator

NEXPTIME-co

D. Della Monica

Extending PNL

PNL


NEXPTIME-co

Undecidable

D. Della Monica

Extending PNL

PNL


MPNL

NEXPTIME-co

Undecidable

D. Della Monica

Extending PNL

PNL


MPNL

NEXPTIME-co

Undecidable

between EXPSPACEand 2NEXPTIME

D. Della Monica

Extending PNL

PNL


MPNL MPNL+

NEXPTIME-co

Undecidable


D. Della Monica

Extending PNL

PNL


MPNL MPNL+

NEXPTIME-co

Undecidable


Undecidable

D. Della Monica

Extending PNL

PNL


MPNL MPNL+

Hybrid Extensions

NEXPTIME-co

Undecidable


Undecidable

D. Della Monica

Extending PNL

PNL


MPNL MPNL+

Hybrid Extensions

NEXPTIME-co

Undecidable


Undecidable

???

D. Della Monica

Possible hybrid extension of PNL and MPNL

Nominals are definable in PNL(Basic Hybrid PNL)

D. Della Monica


Binders over state variables (intervals)(Strongly Hybrid MPNL)

lead to undecidability


D. Della Monica





D. Della Monica




Binders over length of intervals(Weakly Hybrid MPNL)


D. Della Monica

Weakly Hybrid MPNL (WHMPNL)

Metric constraints of MPNL use constants

len=5, len>2, . . .

WHMPNL allows one to store the length of the currentinterval and to refer to it in sub-formulae

↓x (. . . |=|x), ↓x (. . . |≤|x), . . .

D. Della Monica

WHMPNL fragments

Remark

◮ Constant metric constraints are inter-definable

◮ Hybrid metric constraints ARE NOT!!!(e.g.: you cannot define len≤x in terms of len=x)

D. Della Monica

WHMPNL fragments

Remark



Possible choices:

1. which subset of hybrid constraints among {<,≤,=,≥, >}

D. Della Monica

WHMPNL fragments

Remark



Possible choices:


2. constant metric constraints are allowed or not (WHPNL orWHMPNL)

D. Della Monica

WHMPNL fragments

Remark



Possible choices:



3. how many length variables

D. Della Monica

WHMPNL fragments

Remark



Possible choices:




set of hybrid constant # of lengthconstraints constraints variables

WHMPNL(<,≤,=,≥, >) {<,≤,=,≥, >} YES unboundedWHPNL(<,=) {<,=} NO unboundedWHPNL(<)1 {<} NO 1

D. Della Monica

WHMPNL fragments

Remark



Possible choices:




set of hybrid constant # of lengthconstraints constraints variables

WHMPNL(<,≤,=,≥, >) {<,≤,=,≥, >} YES unboundedWHPNL(<,=) {<,=} NO unboundedWHPNL(<)1 {<} NO 1

D. Della Monica

The fragment WHPNL(=)1

Reduction from the Finite Tiling ProblemThis is the problem of establishing whether, for a given finite set of tiletypes T = {t1, . . . , tk}, there exists a finite rectangle R having theborder colored with a fixed color such that T can tile R respecting thecolor constraints.

D. Della Monica

The fragment WHPNL(=)1


D. Della Monica

Outline






D. Della Monica

Extending PNL

PNL


MetricPNL

MPNL+

Hybrid Extensions

NEXPTIME-co

Undecidable


Undecidable

D. Della Monica

Extending PNL

PNL


MetricPNL

MPNL+

Hybrid Extensions

NEXPTIME-co

Undecidable


Undecidable

Undecidable

D. Della Monica

Extending PNL

PNL


MetricPNL

MPNL+

Hybrid Extensions

First-Order ExtensionsNEXPTIME-co

Undecidable


Undecidable

Undecidable

D. Della Monica

Extending PNL

PNL


MetricPNL

MPNL+

Hybrid Extensions


Undecidable


Undecidable

Undecidable

???

D. Della Monica

First-Order together with Propositional

F O R P N L

First-Order Right Propositional Neighborhood Logic

D. Della Monica


F O R P N L


D. Della Monica


F O R P N L


D. Della Monica


F O R P N L


D. Della Monica


F O R P N L


1. Propositional (modal) setting

w1

w2

w4

w3

D. Della Monica


F O R P N L



2. First-Order setting◮ predicates over elements◮ existential and universal

quantificationsw1

w2

w4

w3

D. Della Monica


F O R P N L




quantificationsw1

w4

w3w2

D. Della Monica


F O R P N L




quantificationsw1

w2

w4

w3

D. Della Monica


F O R P N L




quantifications

3. Propositional (modal) +First-Order setting

w1

w2

w3d1

d2

d3

w4

Propositional (modal)domain

First-Orderdomain

D. Della Monica

Parameters of the logic

◮ Temporal domain: discrete, dense, finite, bounded,unbounded, . . .

D. Della Monica



◮ First-order domain: finite, infinite, expanding, . . .

D. Della Monica




◮ First-order constructs:◮ predicates P(. . .),Q(. . .), . . .◮ individual variables x , y , . . .◮ individual constants a, b, . . .◮ function f (. . .), g(. . .), . . .◮ quantifiers◮ terms t1, t2, . . . (variables, constants, and functions)

D. Della Monica





D. Della Monica





D. Della Monica





terms = variables

D. Della Monica





terms = variables

for tight undecidability only 1 variable (no free variables)

D. Della Monica

Undecidability of FORPNL


D. Della Monica



D. Della Monica



It is possible to simulate HS operators 〈B〉〈E〉〈D〉

D. Della Monica

Extending PNL: the final picture

PNL


MetricPNL

MPNL+

Hybrid Extensions


Undecidable


Undecidable

Undecidable

Undecidable

D. Della Monica

Outline






D. Della Monica


This talk outlined several major topics in the area of interval logics:

D. Della Monica



◮ Expressiveness of HS fragments

D. Della Monica




◮ Undecidability of HS fragments

D. Della Monica





◮ Classical extension of PNL

D. Della Monica






The main research agenda so far: to complete the classifications ofexpressiveness and (un)decidability of fragments of HS.

D. Della Monica







Not discussed, and not yet explored, but important:

D. Della Monica








◮ Model checking of Interval logics

D. Della Monica








◮ Model checking of Interval logics

◮ Automata-based techniques for interval logics

D. Della Monica

Exams and attended courses

◮ Exams◮ International Lipari Summer School 2008 on “Algorithms:

Science and Engineering” 14 - 25 July 2008, Lipari; 1.5 Credits◮ “Constraint Programming and NMR Constraints for

Determining Protein Structure”, A. Dovier◮ GAMES Spring School 2009◮ “Systems Biology”, A. Policriti/M. Miculan◮ “Computational Complexity (Complessità computazionale)”, R.

Rizzi◮ “Introduction to Software Configuration Management”, L.

Bendix

◮ Other courses◮ “(Meta-)Modeling with UML and OCL”, M. Gogolla◮ “Data Mining and Mathematical Programming”, P. Serafini◮ “Sistemi Reattivi: automi, logica, algoritmi” (Master Course),

A. Montanari◮ English course for academic purposes (CLAV)

D. Della Monica

Other activities

◮ Summer school◮ International Lipari Summer School 2008 on “Algorithms:

Science and Engineering”◮ GAMES Spring School 2009 (Bertinoro)

◮ Visiting◮ Oct - Dec 2009: University of Murcia - Murcia, Spain (G.

Sciavicco)◮ Sept - Nov 2010: Technical University of Denmark (DTU) -

Lyngby, Copenhagen, Denmark (V. Goranko)

◮ Events organization◮ Annual Workshop of the ESF Networking Programme on

Games for Design and Verification (GAMES 2009)◮ First International Symposium on Games, Automata, Logics

and Formal Verification (GandALF 2010)◮ Second International Symposium on Games, Automata, Logics

and Formal Verification (GandALF 2011)

D. Della Monica

Publications

1. D. Bresolin, D. Della Monica, V. Goranko, A. Montanari, and G.

Sciavicco. “Decidable and Undecidable Fragments of Halpernand Shohams Interval Temporal Logic: Towards a CompleteClassification”. In Proc. of 15th International Conference on Logic

for Programming, Artificial Intelligence, and Reasoning (LPAR

2008), 2008.


Sciavicco. “Undecidability of Interval Temporal Logics with theOverlap Modality”. In Proc. of 16th International Symposium on

Temporal Representation and Reasoning (TIME 2009), 2009.


Sciavicco. “Undecidability of the Logic of Overlap Relation overDiscrete Linear Orderings”. Electronic Notes in Theoretical

Computer Science (Proc. of the 6th Workshop on Methods for

Modalities (M4M-6 2009), 2010.

D. Della Monica

Publications - contd’

4. D. Della Monica, V. Goranko, and G. Sciavicco. “Hybrid MetricPropositional Neighborhood Logics with Interval LengthBinders”. In Proc. of International Workshop on Hybrid Logic and

Applications (HyLo 2010), 2010. To appear on ENTCS.

5. D. Della Monica and G. Sciavicco. “On First-Order PropositionalNeighborhood Logics: a First Attempt”. In Proc. of ECAI 2010

Workshop on Spatio-Temporal Dynamics (STeDY 2010), 2010.


Sciavicco. “Metric Propositional Neighborhood Logics:Expressiveness, Decidability, and Undecidability”. In Proc. of

19th European Conference on Artificial Intelligence (ECAI 2010),

2010.

7. D. Bresolin, D. Della Monica, A. Montanari, P. Sala, and G.

Sciavicco. “A decidable spatial generalization of Metric IntervalTemporal Logic”. In Proc. of 17th International Symposium on

Temporal Representation and Reasoning (TIME 2010), 2010.

D. Della Monica


8. D. Bresolin, D. Della Monica, V. Goranko, A. Montanari, G.

Sciavicco. “Metric propositional neighborhood logics on naturalnumbers”. Journal of Software & Systems Modeling (doi:

10.1007/s10270-011-0195-y, online since February 2011).

9. D. Della Monica, V. Goranko, A. Montanari, G. Sciavicco.

“Expressiveness of the Interval Logics of Allen’s Relations onthe Class of all Linear Orders: Complete Classification”.accepted to IJCAI 2011.

D. Della Monica


8. D. Bresolin, D. Della Monica, V. Goranko, A. Montanari, G.

Sciavicco. “Metric propositional neighborhood logics on naturalnumbers”. Journal of Software & Systems Modeling (doi:

10.1007/s10270-011-0195-y, online since February 2011).

9. D. Della Monica, V. Goranko, A. Montanari, G. Sciavicco.

“Expressiveness of the Interval Logics of Allen’s Relations onthe Class of all Linear Orders: Complete Classification”.accepted to IJCAI 2011.

The end.

Documents

Expressiveness, decidability, and undecidability of ...users.dimi.uniud.it/.../thesis/phd_thesis_slides.pdf · D. Della Monica University of Udine Department of Mathematics and Computer