Presentation is loading. Please wait.

Presentation is loading. Please wait.

Alternating tree Automata and Parity games

Similar presentations


Presentation on theme: "Alternating tree Automata and Parity games"โ€” Presentation transcript:

1 Alternating tree Automata and Parity games
Based on chapter 9 Logic and infinite games, edited by Gradel, Thomas and Wilke, LNCS 2500. Lior Schneider,

2 topics Defining Alternating tree automata and its behavior
Parity games equality to Ata The word problem complexity The complementation problem complexity The Emptiness problem complexity

3 perliminaries Through all of this lecture we will fix a set ofย Propositional variables ๐‘ƒ. Propositional variables are theย atomic formulasย of propositional logic Transition system: ๐‘†,๐‘…,๐œ† : ๐‘† : the states ๐‘…โŠ†๐‘†๐‘ฅ๐‘† : relations ๐œ†:๐‘ƒโ†’๐’ซ(๐‘†) : assigns a set of states to every propositional variable Transition systems are known as Kripke structure: ๐‘€=(๐‘†,๐ผ,๐‘…,๐ฟ) ๐‘ƒ= ๐‘,๐‘ž ; ๐‘†= ๐‘  1 , ๐‘  2 , ๐‘  3 ;๐ผ= ๐‘  1 ; ๐‘…= ๐‘  1 , ๐‘  2 , ๐‘  2 , ๐‘  1 , ๐‘  2 , ๐‘  3 , ๐‘  3 , ๐‘  3 ๐ฟ= ๐‘  1 , ๐‘,๐‘ž , ๐‘  2 , ๐‘ž , ๐‘  3 , ๐‘ M produces the path: ๐‘  1 ๐‘  2 ๐‘  1 ๐‘  2 ๐‘  3 ๐‘  3 .. The word: {๐‘,๐‘ž},{๐‘ž},{๐‘,๐‘ž},{๐‘ž},{๐‘},{๐‘},{๐‘},โ€ฆ ๐‘  1 {๐‘,๐‘ž} ๐‘  2 {๐‘ž} ๐‘  3 {๐‘}

4 perliminaries ๐‘  1 {๐‘,๐‘ž} ๐‘  2 {๐‘ž} Transition system:
For every variable ๐‘โˆˆ๐‘ƒ and every state ๐‘ โˆˆ๐œ†(๐‘), we say that ๐‘ is true in ๐‘ , and for ๐‘ โˆˆ๐‘†\ ๐œ†(๐‘), we say that ๐‘ is false in ๐‘ . ๐‘“๐‘œ๐‘Ÿ ๐‘’๐‘Ž๐‘โ„Ž ๐‘ โˆˆ๐‘†, ๐‘ ๐‘… = {๐‘ โ€ฒ โˆˆ ๐‘† | (๐‘ , ๐‘ โ€ฒ) โˆˆ ๐‘…} ๐‘Ž๐‘›๐‘‘ ๐‘…๐‘  = {๐‘ โ€ฒ โˆˆ ๐‘† | (๐‘ โ€ฒ, ๐‘ ) โˆˆ ๐‘…}. Pointed transition system: (๐’ฎ, ๐‘  ๐ผ ) We call a transition system ๐’ฎ= (๐‘†,๐‘…, ๐œ†) (๐‘Ÿ๐‘’๐‘ ๐‘ (๐’ฎ, ๐‘  ๐ผ )) finite iff ๐‘† is finite and ๐œ†(๐‘) =โˆ… for just finitely many ๐‘โˆˆ๐‘ƒ. ๐‘  3 {๐‘}

5 perliminaries transition conditions ๐‘‡ ๐ถ ๐‘„ ๐‘œ๐‘ฃ๐‘’๐‘Ÿ ๐‘„ ๐‘ ๐‘’๐‘ก ๐‘œ๐‘“ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’๐‘  :
The symbols 0 and 1 are transition condition For every ๐‘โˆˆ๐‘ƒ, ๐‘ and ยฌ๐‘ are transition conditions. For every ๐‘žโˆˆ๐‘„, ๐‘ž, โˆŽ๐‘ž, โˆ†๐‘ž are transition conditions. For every ๐‘ž 1 , ๐‘ž 2 โˆˆ๐‘„, ๐‘ž 1 โˆง ๐‘ž 2 ๐‘Ž๐‘›๐‘‘ ๐‘ž 1 โˆจ ๐‘ž 2 are transition conditions. Note that this definition does not allow transition conditions like ๐‘ž 1 โˆงโˆŽ ๐‘ž 2 ๐‘œ๐‘Ÿ pโˆง q for p โˆˆ P and q โˆˆ Q.

6 perliminaries Alternating Tree Automata: ๐’œ= ๐‘„, ๐‘ž ๐ผ , ๐›ฟ,ฮฉ
๐’œ= ๐‘„, ๐‘ž ๐ผ , ๐›ฟ,ฮฉ ๐‘„ is a finite set of states of the automaton. ๐‘ž ๐ผ โˆˆ๐‘„ is a state called the initial state. ๐›ฟ :๐‘„โ†’ ๐‘‡ ๐ถ ๐‘„ is called transition function. ฮฉ :๐‘„โ†’๐œ” is called priority function.

7 Alternating Tree Automata behavior
An instance of Ata is a tuple ๐‘ž,๐‘  ๐œ–๐‘„ร—๐‘†. The behavior of an Ata defined by the current instance. The automaton operates ๐›ฟ ๐‘ž ๐‘Ž๐‘›๐‘‘ ๐‘Ž๐‘๐‘ก๐‘  ๐‘Ž๐‘๐‘๐‘œ๐‘Ÿ๐‘‘๐‘–๐‘›๐‘”๐‘™๐‘ฆ: (๐‘ž ๐ผ ,๐‘  ๐ผ ) ๐›ฟ ๐‘ž ๐ผ = 0,1 , ๐‘, ยฌ๐‘ ๐›ฟ ๐‘ž ๐ผ =โˆŽ/โˆ†๐‘žโ€ฒโˆˆ๐‘„ stops (๐‘žโ€ฒ,๐‘  1 ) ๐›ฟ ๐‘ž ๐ผ = ๐‘ž โ€ฒ , ๐‘žโ€ฒโˆˆ๐‘„ ๐‘… ๐‘š๐‘Ž๐‘›๐‘ฆ ๐‘ก๐‘–๐‘š๐‘’๐‘  ๐›ฟ ๐‘ž ๐ผ = ๐‘ž 1 โˆง/โˆจ ๐‘ž 2 ; ๐‘ž 1 , ๐‘ž 2 โˆˆ๐‘„ (๐‘žโ€ฒ,๐‘  3 ) (๐‘žโ€ฒ,๐‘  ๐ผ ) (๐‘žโ€ฒ,๐‘  2 ) ( ๐‘ž 2 ,๐‘  ๐ผ ) ( ๐‘ž 1 ,๐‘  ๐ผ )

8 successful instance ๐‘  1 {๐‘,๐‘ž} ๐‘  2 {๐‘ž} ๐‘  3 {๐‘}
We define a successful instance ๐‘ž,๐‘  accordingly: Recall: For every variable ๐‘โˆˆ๐‘ƒ and every state ๐‘ โˆˆ๐œ†(๐‘), we say that ๐‘ is true in ๐‘ , and for ๐‘ โˆˆ๐‘†\ ๐œ†(๐‘), we say that ๐‘ is false in ๐‘ . ๐›ฟ ๐‘ž =๐‘ ๐‘Ž๐‘›๐‘‘ ๐‘ ๐‘–๐‘  ๐‘ก๐‘Ÿ๐‘ข๐‘’ ๐‘–๐‘› ๐‘  (๐‘โˆˆ๐œ†(๐‘ )) ๐›ฟ ๐‘ž =ยฌ๐‘ ๐‘Ž๐‘›๐‘‘ ๐‘ ๐‘–๐‘  ๐‘“๐‘Ž๐‘™๐‘ ๐‘’ ๐‘–๐‘› ๐‘  (๐‘โˆ‰๐œ†(๐‘ )) ๐›ฟ ๐‘ž =1 If ๐›ฟ(๐‘ž) =๐‘žโ€ฒ The instance (๐‘ž, ๐‘ ) is successful iff (๐‘žโ€ฒ, ๐‘ ) is successful. If ๐›ฟ(๐‘ž) = ๐‘ž 1 โˆง ๐‘ž 2 , then the instance (๐‘ž, ๐‘ ) succeeds iff both the instances ( ๐‘ž 1 , ๐‘ ) and ( ๐‘ž 2 , ๐‘ ) succeed. If ๐›ฟ(๐‘ž)= ๐‘ž 1 โˆจ ๐‘ž 2 , then the instance succeeds iff at least one of the instances ( ๐‘ž 1 , ๐‘ ) and ( ๐‘ž 2 , ๐‘ ) succeeds. If ๐›ฟ(๐‘ž) =โˆŽ๐‘žโ€ฒ, then the instance (๐‘ž, ๐‘ ) succeeds iff for every ๐‘ โ€ฒโˆˆ๐‘ ๐‘… the instance (๐‘žโ€ฒ, ๐‘ โ€ฒ) succeed. If ๐›ฟ(๐‘ž) =โˆ†๐‘žโ€ฒ, then the instance (๐‘ž, ๐‘ ) succeeds iff for at least one ๐‘ โ€ฒโˆˆ๐‘ ๐‘… the instance (๐‘žโ€ฒ, ๐‘ โ€ฒ) succeed. Finally: The automaton accepts the transition system (๐“ข, ๐’” ๐‘ฐ ) iff the initial instance ( ๐’’ ๐‘ฐ , ๐’” ๐‘ฐ ) succeeds.

9 Formalization problem
If we have an infinite tree, how can we determine if an instance is successful or not? ๐›ฟ ๐‘ž =๐‘ž We will convert the automaton into a parity game! (๐‘ž,๐‘ ) ๐›ฟ ๐‘ž =๐‘ž

10 Parity game conversion
Let ๐’œ= ๐‘„, ๐‘ž ๐ผ ,๐›ฟ,ฮฉ , ๐’ฎ= ๐‘†, ๐‘  ๐‘– , we will examine a word ๐‘ฃโˆˆ ๐‘„ร—๐‘† โˆ— and a letter ๐‘ž,๐‘  โˆˆ๐‘„ร—๐‘†. (notation of ๐‘ฃ(๐‘ž,๐‘ ) word concatenation with letter) We will define ๐’œ on ๐’ฎ behavior as the least language ๐‘‰โŠ† ๐‘„ร—๐‘† โˆ— such that ๐‘ž ๐ผ , ๐‘  ๐ผ โˆˆ๐‘‰ and โˆ€๐‘ฃ ๐‘ž,๐‘  โˆˆ๐‘‰: ๐›ฟ ๐‘ž = ๐‘ž โ€ฒ โ†’๐‘ฃ ๐‘ž,๐‘  ๐‘ž โ€ฒ ,๐‘  โˆˆ๐‘‰ ๐›ฟ ๐‘ž = ๐‘ž 1 โˆง ๐‘ž 2 ๐‘œ๐‘Ÿ ๐‘ž 1 โˆจ ๐‘ž 2 โ†’๐‘ฃ ๐‘ž,๐‘  ๐‘ž 1 ,๐‘  โˆˆ๐‘‰ and ๐‘ฃ ๐‘ž,๐‘  ๐‘ž 2 ,๐‘  โˆˆ๐‘‰ ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ ๐‘œ๐‘Ÿ โˆ†๐‘žโ€ฒโ†’๐‘ฃ ๐‘ž,๐‘  ๐‘žโ€ฒ,๐‘ โ€ฒ โˆˆ๐‘‰

11 Parity game conversion
The Arena ๐‘‰ 0 , ๐‘‰ 1 ,๐ธ : Moves ๐ธโ‰œ ๐‘ฃ, ๐‘ฃ ๐‘ž,๐‘  ๐‘ฃ ๐‘ž,๐‘  โˆˆ๐‘‰ } a word and concatenated word such that the concatenated word is in V. The parity game priority mapping function will be ฮฉ ๐‘ฃ ๐‘ž,๐‘  โ‰œฮฉ ๐‘ž โˆ€๐‘ฃ ๐‘ž,๐‘  โˆˆ๐‘‰ ๐’œ ๐‘Ž๐‘๐‘๐‘’๐‘๐‘ก๐‘  ๐’ฎ ๐‘–๐‘“๐‘“ โˆƒ winning strategy for player 0 in the equivalent parity game ๐’ข=( ๐‘‰ 0 , ๐‘‰ 1 ,๐ธ , ฮฉ, ๐‘ž ๐ผ , ๐‘  ๐ผ ) โ„’ ๐’œ =๐‘Ž๐‘™๐‘™ ๐‘กโ„Ž๐‘’ ๐‘๐‘œ๐‘–๐‘›๐‘ก๐‘’๐‘‘ ๐‘ก๐‘Ÿ๐‘Ž๐‘›๐‘ ๐‘–๐‘ก๐‘–๐‘œ๐‘› ๐‘ ๐‘ฆ๐‘ ๐‘ก๐‘’๐‘š ๐‘กโ„Ž๐‘Ž๐‘ก ๐’œ ๐‘Ž๐‘๐‘๐‘’๐‘๐‘ก๐‘  ๐‘ฝ ๐Ÿ ๐‘ฝ ๐ŸŽ ๐›ฟ ๐‘ž =1 ๐›ฟ ๐‘ž =0 ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘žโ€ฒ ๐›ฟ ๐‘ž = ๐‘ž 1 โˆง ๐‘ž 2 ๐›ฟ ๐‘ž = ๐‘ž 1 โˆจ ๐‘ž 2 ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ ๐›ฟ ๐‘ž =ฮ”๐‘žโ€ฒ

12 Examples ๐‘„= ๐‘ž ๐ผ ; ๐›ฟ ๐‘ž ๐ผ = โˆŽ ๐‘ž ๐ผ ;ฮฉ ๐‘ž ๐ผ =0 ๐‘คโ„Ž๐‘’๐‘Ÿ๐‘’ ๐’ฎ ๐‘–๐‘  ๐‘Ž๐‘›๐‘ฆ ๐‘๐‘œ๐‘–๐‘›๐‘ก๐‘’๐‘‘ ๐‘ก๐‘Ÿ๐‘Ž๐‘›๐‘ ๐‘–๐‘ก๐‘–๐‘œ๐‘› ๐‘ ๐‘ฆ๐‘ ๐‘ก๐‘’๐‘š. Priority, loacation Notice that player 0 doesnโ€™t have any Loacation but the only priority is 0 Hence player 1 looses for every game (finite and infinite), therefor the Automaton accepts any pointed transition system. 0,1 (๐‘ž ๐ผ ,๐‘  ๐ผ ) ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒโˆˆ๐‘„ 0,1 ( ๐‘ž ๐ผ ,๐‘  1 ) ๐‘… ๐‘š๐‘Ž๐‘›๐‘ฆ ๐‘ก๐‘–๐‘š๐‘’๐‘  0,1 ( ๐‘ž ๐ผ ,๐‘  3 ) 0,1 ( ๐‘ž ๐ผ ,๐‘  2 ) ๐‘ฝ ๐Ÿ ๐‘ฝ ๐ŸŽ ๐›ฟ ๐‘ž =1 ๐›ฟ ๐‘ž =0 ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘žโ€ฒ ๐›ฟ ๐‘ž = ๐‘ž 1 โˆง ๐‘ž 2 ๐›ฟ ๐‘ž = ๐‘ž 1 โˆจ ๐‘ž 2 ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ ๐›ฟ ๐‘ž =ฮ”๐‘žโ€ฒ

13 Examples ๐‘„= ๐‘ž ๐ผ ; ๐›ฟ ๐‘ž ๐ผ = โˆŽ ๐‘ž ๐ผ ;ฮฉ ๐‘ž ๐ผ =1 where ๐’ฎ is any pointed transition system with some infinite path starting at ๐‘  ๐ผ . Priority, loacation Notice that player 0 still doesnโ€™t have any Loacation but the only priority is 1 Hence player 1 wins for every infinite Game by choosing the infinite path If ๐’ฎ is any pointed transition system with no infinite path, player 1 will loose. therefor the Automaton accepts any pointed transition system with no infinite paths. 1,1 (๐‘ž ๐ผ ,๐‘  ๐ผ ) ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒโˆˆ๐‘„ 1,1 ( ๐‘ž ๐ผ ,๐‘  1 ) ๐‘… ๐‘š๐‘Ž๐‘›๐‘ฆ ๐‘ก๐‘–๐‘š๐‘’๐‘  1,1 ( ๐‘ž ๐ผ ,๐‘  3 ) 1,1 ( ๐‘ž ๐ผ ,๐‘  2 ) ๐‘ฝ ๐Ÿ ๐‘ฝ ๐ŸŽ ๐›ฟ ๐‘ž =1 ๐›ฟ ๐‘ž =0 ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘žโ€ฒ ๐›ฟ ๐‘ž = ๐‘ž 1 โˆง ๐‘ž 2 ๐›ฟ ๐‘ž = ๐‘ž 1 โˆจ ๐‘ž 2 ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ ๐›ฟ ๐‘ž =ฮ”๐‘žโ€ฒ

14 Examples ๐‘„= ๐‘ž ๐ผ ; ๐›ฟ ๐‘ž ๐ผ =ฮ” ๐‘ž ๐ผ ;ฮฉ ๐‘ž ๐ผ =1 ๐‘คโ„Ž๐‘’๐‘Ÿ๐‘’ ๐’ฎ ๐‘–๐‘  ๐‘Ž๐‘›๐‘ฆ ๐‘๐‘œ๐‘–๐‘›๐‘ก๐‘’๐‘‘ ๐‘ก๐‘Ÿ๐‘Ž๐‘›๐‘ ๐‘–๐‘ก๐‘–๐‘œ๐‘› ๐‘ ๐‘ฆ๐‘ ๐‘ก๐‘’๐‘š. Priority, loacation All the location are player 0 hence Player 0 can choose a finite path and therefore the automaton Doesnโ€™t accept any ๐’ฎ 1,0 (๐‘ž ๐ผ ,๐‘  ๐ผ ) ๐›ฟ ๐‘ž =ฮ”๐‘žโ€ฒโˆˆ๐‘„ 1,0 ( ๐‘ž ๐ผ ,๐‘  1 ) ๐‘… ๐‘š๐‘Ž๐‘›๐‘ฆ ๐‘ก๐‘–๐‘š๐‘’๐‘  1,0 ( ๐‘ž ๐ผ ,๐‘  3 ) ๐‘ฝ ๐Ÿ ๐‘ฝ ๐ŸŽ ๐›ฟ ๐‘ž =1 ๐›ฟ ๐‘ž =0 ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆ‰๐œ†(๐‘) ๐›ฟ ๐‘ž =ยฌ๐‘โˆง๐‘ โˆˆ๐œ†(๐‘) ๐›ฟ ๐‘ž =๐‘žโ€ฒ ๐›ฟ ๐‘ž = ๐‘ž 1 โˆง ๐‘ž 2 ๐›ฟ ๐‘ž = ๐‘ž 1 โˆจ ๐‘ž 2 ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ ๐›ฟ ๐‘ž =ฮ”๐‘žโ€ฒ 1,0 ( ๐‘ž ๐ผ ,๐‘  2 )

15 Current disadvantages
While ๐’ฎ may be finite our parity game ๐’ข may be infinite (for example ๐›ฟ ๐‘ž ๐ผ = ๐‘ž ๐ผ ) therefor we have the need to define the behavior in more convenient way. This would be helpful for our next slides

16 Alternative Formal Definition for the Arena
The same ๐’œ= ๐‘„, ๐‘ž ๐ผ ,๐›ฟ,ฮฉ , ๐’ฎ= ๐‘†, ๐‘  ๐‘– , Let ๐‘‰ โŠ†๐‘„ร—๐‘† and E โŠ† ๐‘‰ ร— ๐‘‰ be the smallest graph with ๐‘ž ๐ผ , ๐‘  ๐ผ โˆˆ ๐‘‰ ๐‘ ๐‘ข๐‘โ„Ž ๐‘กโ„Ž๐‘Ž๐‘ก โˆ€ ๐‘ž,๐‘  โˆˆ ๐‘‰ ๐‘ค๐‘’ โ„Ž๐‘Ž๐‘ฃ๐‘’: ๐›ฟ ๐‘ž = ๐‘ž โ€ฒ โ†’ ๐‘ž โ€ฒ ,๐‘  โˆˆ[๐‘‰] ๐‘Ž๐‘›๐‘‘ ๐‘ž,๐‘  ๐‘ž โ€ฒ ,๐‘  โˆˆ[๐ธ] ๐›ฟ ๐‘ž = ๐‘ž 1 โˆง ๐‘ž 2 ๐‘œ๐‘Ÿ ๐‘ž 1 โˆจ ๐‘ž 2 โ†’ ๐‘ž 1 ,๐‘  , ๐‘ž 2 ,๐‘  โˆˆ[๐‘‰] and ๐‘ž,๐‘  , ๐‘ž 1 ,๐‘  , ๐‘ž,๐‘  ๐‘ž 2 ,๐‘  โˆˆ[๐ธ] ๐›ฟ ๐‘ž =โˆŽ ๐‘ž โ€ฒ ๐‘œ๐‘Ÿ โˆ† ๐‘ž โ€ฒ โ†’ ๐‘ž โ€ฒ , ๐‘  โ€ฒ โˆˆ ๐‘‰ ๐‘Ž๐‘›๐‘‘ ๐‘ž,๐‘  , ๐‘ž โ€ฒ , ๐‘  โ€ฒ โˆˆ ๐ธ โˆ€ ๐‘  โ€ฒ โˆˆsR We split ๐‘‰ ๐‘–๐‘›๐‘ก๐‘œ ๐‘‰ 0 ๐‘Ž๐‘›๐‘‘ ๐‘‰ 1 ๐‘Ž๐‘  ๐‘’๐‘Ž๐‘Ÿ๐‘™๐‘–๐‘’๐‘Ÿ ๐‘Ž๐‘›๐‘‘ ๐‘กโ„Ž๐‘’ ฮฉ ๐‘–๐‘  ๐‘กโ„Ž๐‘’ ๐‘ ๐‘Ž๐‘š๐‘’. Previous definition: We will define ๐’œ on ๐’ฎ behavior as the least language ๐‘‰โŠ† ๐‘„ร—๐‘† โˆ— such that ๐‘ž ๐ผ , ๐‘  ๐ผ โˆˆ๐‘‰ and โˆ€๐‘ฃ ๐‘ž,๐‘  โˆˆ๐‘‰: ๐›ฟ ๐‘ž = ๐‘ž โ€ฒ โ†’๐‘ฃ ๐‘ž,๐‘  ๐‘ž โ€ฒ ,๐‘  โˆˆ๐‘‰ ๐›ฟ ๐‘ž = ๐‘ž 1 โˆง ๐‘ž 2 ๐‘œ๐‘Ÿ ๐‘ž 1 โˆจ ๐‘ž 2 โ†’๐‘ฃ ๐‘ž,๐‘  ๐‘ž 1 ,๐‘  โˆˆ๐‘‰ and ๐‘ฃ ๐‘ž,๐‘  ๐‘ž 2 ,๐‘  โˆˆ๐‘‰ ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ ๐‘œ๐‘Ÿ โˆ†๐‘žโ€ฒโ†’๐‘ฃ ๐‘ž,๐‘  ๐‘žโ€ฒ,๐‘ โ€ฒ โˆˆ๐‘‰

17 Alternative Formal Definition for the Arena
Our new Arena: ๐’ข= ๐‘‰ 0 , ๐‘‰ 1 , ๐ธ , ฮฉ, ๐‘ž ๐ผ , ๐‘  ๐ผ . Definition of the function []: ๐‘„ร—S + โ†’(๐‘„ร—๐‘†) for each word ๐‘„ร—S + it assigns the value of its last letter ๐‘ž,๐‘  Note that [] maps location from V to [V], preserve edges and priorities, therefor ๐’ขโ€ฒ is created by applying [] on ๐’ข

18 Alternative Formal Definition for the Arena
Theorem: Player 0 has a winning strategy in ๐’ข iff Player 0 has a winning strategy in ๐’ขโ€˜ Proof: Let ๐‘“โ€ฒ 0 :[ ๐‘‰ 0 ]โ†’[๐‘‰] be a winning strategy for Player 0 in ๐’ขโ€ฒ We define a mapping ๐‘“ 0 : ๐‘‰ 0 โ†’๐‘‰ ๐‘๐‘ฆ ๐‘“ 0 ๐‘ฃ โ‰œ ๐‘ฃ ๐‘“ 0 โ€ฒ ๐‘ฃ . Let ๐œ‹ be a play in ๐’ข which is consistent with ๐‘“ 0 . Then, [๐œ‹] is consistent with ๐‘“โ€ฒ 0 , and thus, [๐œ‹] and ๐œ‹ are won by Player 0. Consequently, ๐‘“ 0 is a winning strategy for Player 0 in ๐’ข. Clearly, we can apply a symmetric argument if ๐‘“ 1 โ€ฒ :[ ๐‘‰ 1 ]โ†’[๐‘‰] is a winning strategy for Player 1 in ๐’ขโ€ฒ.

19 Complex Transition Conditions
Recall: transition conditions ๐‘‡ ๐ถ ๐‘„ ๐‘œ๐‘ฃ๐‘’๐‘Ÿ ๐‘„ ๐‘ ๐‘’๐‘ก ๐‘œ๐‘“ ๐‘ ๐‘ก๐‘Ž๐‘ก๐‘’๐‘  : The symbols 0 and 1 are transition condition For every ๐‘โˆˆ๐‘ƒ, ๐‘ and ยฌ๐‘ are transition conditions. For every ๐‘žโˆˆ๐‘„, ๐‘ž, โˆŽ๐‘ž, โˆ†๐‘ž are transition conditions. For every ๐‘ž 1 , ๐‘ž 2 โˆˆ๐‘„, ๐‘ž 1 โˆง ๐‘ž 2 ๐‘Ž๐‘›๐‘‘ ๐‘ž 1 โˆจ ๐‘ž 2 are transition conditions. Note that this definition does not allow transition conditions like ๐‘ž 1 โˆงโˆŽ ๐‘ž 2 ๐‘œ๐‘Ÿ pโˆง q for p โˆˆ P and q โˆˆ Q. Example of such ๐œ‘: โ€œChange the inner state to ๐‘ž 1 if ๐‘ is true, otherwise change the inner state to ๐‘ž 2 โ€ Formally: ๐œ‘=( ๐‘ž 1 โˆง๐‘)โˆจ( ๐‘ž 2 โˆง๏ฟข๐‘) So we will use some new states: ๐‘ž ๐œ‘ , ๐‘ž ๐‘ž 1 โˆง๐‘ , ๐‘ž ๐‘ž 1 โˆงยฌ๐‘ , ๐‘ž ๐‘ , ๐‘ž ยฌ๐‘

20 Complex Transition Conditions
(๐‘ž ๐ผ ,๐‘  ๐ผ ) 1,0 ๐›ฟ ๐‘ž ๐œ‘ โ‰œ ๐‘ž ๐‘ž 1 โˆง๐‘ โˆจ ๐‘ž ๐‘ž 2 โˆงยฌ๐‘ ๐›ฟ ๐‘ž ๐‘ž 1 โˆง๐‘ โ‰œ ๐‘ž 1 โˆง ๐‘ž ๐‘ ๐›ฟ ๐‘ž ๐‘ž 2 โˆงยฌ๐‘ โ‰œ ๐‘ž 2 โˆง ๐‘ž ยฌ๐‘ ๐›ฟ ๐‘ž ๐‘ โ‰œ๐‘ ๐›ฟ ๐‘ž ยฌ๐‘ โ‰œยฌ๐‘ Example: We set: ๐›ฟ ๐‘ž 1 =๐›ฟ ๐‘ž 2 =โˆŽ ๐‘ž ๐œ‘ ;ฮฉ ๐‘ž 1 =2;ฮฉ ๐‘„\ q 1 =1; ๐‘ž ๐ผ = ๐‘ž ๐œ‘ Priority, loacation The automaton accepts some pointed transition system ๐’ฎ iff every infinite path starting from ๐‘  ๐ผ contains infinitely many states in which p is true. There will be infinite amount of successful ๐‘ž 1 , ๐‘  โ€ฒ ๐‘–๐‘›๐‘ ๐‘ก๐‘Ž๐‘›๐‘๐‘’๐‘  ๐‘ค๐‘–๐‘กโ„Ž ๐‘’๐‘ฃ๐‘’๐‘› ๐‘๐‘Ÿ๐‘–๐‘œ๐‘Ÿ๐‘–๐‘ก๐‘ฆ. ๐›ฟ ๐‘ž ๐œ‘ = ๐‘ž ๐‘ž 1 โˆง๐‘ โˆจ ๐‘ž ๐‘ž 2 โˆงยฌ๐‘ 1,1 1,1 ( ๐‘ž ๐‘ž 1 โˆง๐‘ ,๐‘  ๐ผ ) ( ๐‘ž ๐‘ž 2 โˆงยฌ๐‘ ,๐‘  ๐ผ ) 2,1 1,? 1,1 1,1 (๐‘ž 1 ,๐‘  ๐ผ ) ( ๐‘ž ๐‘ ,๐‘  ๐ผ ) (๐‘ž 2 ,๐‘  ๐ผ ) ( ๐‘ž ยฌ๐‘ ,๐‘  ๐ผ ) 1,0 1,0 1,0 1,0 (๐‘ž ๐œ‘ ,๐‘ โ€ฒ 1 ) (๐‘ž ๐œ‘ ,๐‘ โ€ฒ 2 ) (๐‘ž ๐œ‘ ,๐‘ โ€ฒ 2 ) p (๐‘ž ๐œ‘ ,๐‘ โ€ฒ 1 ) ยฌp

21 The Word problem Means to decide whether a given alternating tree automaton ๐’œ accepts a given finite pointed transition system ๐’ฎ. Motivation: In Section 10, this result will be used to determine the complexity of the model checking problem for the modal ฮผ-calculus. (which we will not elaborate on) A naรฏve approach to solve the word problem would be to compute the whole behavior of the Ata. This is actually impossible to solve it that way because it can be infinite behavior. Luckily we learnt some tools to deal with it. We will reduce the equivalent parity game into a finite one which will be a great help for us

22 The Word problem Theorem: Let ๐’œ=(๐‘„, ๐‘ž ๐ผ , ๐›ฟ,ฮฉ) be an alternating tree automaton with d different non-zero priorities and let ๐’ฎ be a finite pointed transition system: There is an algorithm which computes in time: ๐’ช ๐‘‘ ๐‘„ ๐‘… ๐‘„ ๐‘† +1 ๐‘‘ ๐‘‘ 2 and in space ๐’ช ๐‘‘ ๐‘„ ๐‘† log ๐‘„ ๐‘† The word problem is in ๐‘ˆ๐‘ƒโˆฉ๐‘๐‘œโˆ’๐‘ˆ๐‘ƒ

23 The Word problem The word problem is in ๐‘ˆ๐‘ƒโˆฉ๐‘๐‘œโˆ’๐‘ˆ๐‘ƒ:
ย UPย ("Unambiguous Non-deterministic Polynomial-time") decision problemsย solvable inย polynomial timeย on aย unambiguous Turing machineย with at most one accepting path for each input. ๐‘ƒโŠ†๐‘ˆ๐‘ƒโŠ†๐‘๐‘ƒ A language is inย UPย if a given answer can be verified in polynomial time, and the verifier machine only accepts at mostย oneย answer for each problem instance. UP contains parity gameย problem and specifically the question weather player 0 has a winning strategy is in ๐‘ˆ๐‘ƒโˆฉ๐‘๐‘œโˆ’๐‘ˆ๐‘ƒ as explained in chapter 6.

24 The Word problem There is an algorithm which computes in time: ๐’ช ๐‘‘ ๐‘„ ๐‘… ๐‘„ ๐‘† +1 ๐‘‘ ๐‘‘ 2 and in space ๐’ช ๐‘‘ ๐‘„ ๐‘† log ๐‘„ ๐‘† Upper bounds: Lets fix ๐’œ and examine ๐’ฎ complexity: ๐‘… โ‰ค ๐‘† 2 ;๐‘‘ ๐‘„ ๐‘Ž๐‘Ÿ๐‘’ ๐’œ ๐‘๐‘œ๐‘›๐‘ ๐‘ก๐‘Ž๐‘›๐‘ก there for we get:๐’ช ๐‘† 2 ( ๐‘„ ๐‘† ๐‘‘ ) Roughly we get ๐’ช ( ๐‘† 2+ ๐‘‘ ) With ๐‘‘=2 we get ๐‘† 3 , with ๐‘‘=20 we get |๐‘†|^12 โ€“ not so nice

25 The Word problem There is an algorithm which computes in time: ๐’ช ๐‘‘ ๐‘„ ๐‘… ๐‘„ ๐‘† +1 ๐‘‘ ๐‘‘ 2 and in space ๐’ช ๐‘‘ ๐‘„ ๐‘† log ๐‘„ ๐‘† Lets estimate ๐ธ ๐‘Ž๐‘›๐‘‘ | ๐‘‰ |: As we know, ๐‘‰ โŠ†๐‘„ร—๐‘†โ†’ ๐‘‰ โ‰ค|๐‘„||๐‘†| Let ๐‘† โ€ฒ โŠ†๐‘† be the set of states in ๐’ฎ which are reachable from ๐‘ž ๐ผ Every state in ๐‘† except ๐‘  ๐ผ has at least one predecessor ( ๐‘  ๐ผ ๐‘š๐‘Ž๐‘ฆ โ„Ž๐‘Ž๐‘ฃ๐‘’ ๐‘›๐‘œ ๐‘๐‘Ÿ๐‘’), Hence, ๐‘… โ‰ฅ ๐‘† โˆ’1โ†’ ๐‘… +1โ‰ฅ ๐‘† To determine |[๐ธ]|, we count the number of successors of every location in |[๐‘‰]|. The successors of a location (๐‘ž, ๐‘ )โˆˆ[๐‘‰] are (๐‘ž, ๐‘ )[๐ธ]. ๐ธ = ๐‘žโˆˆ๐‘„ ๐‘ž,๐‘  โˆˆ[๐‘‰] | ๐‘ž,๐‘  ๐ธ | For a fixed ๐‘ž:

26 The Word problem ๐ธ = ๐‘žโˆˆ๐‘„ ๐‘ž,๐‘  โˆˆ[๐‘‰] | ๐‘ž,๐‘  ๐ธ | For a fixed ๐‘ž: To sum up:
๐ธ = ๐‘žโˆˆ๐‘„ ๐‘ž,๐‘  โˆˆ[๐‘‰] | ๐‘ž,๐‘  ๐ธ | For a fixed ๐‘ž: If ๐›ฟ(๐‘ž)โˆˆ{0, 1} ๐‘œ๐‘Ÿ ๐›ฟ(๐‘ž)โˆˆ{๐‘, ๏ฟข๐‘}, then (๐‘ž, ๐‘ ) has no successor: ๐‘ž,๐‘  โˆˆ[๐‘‰] | ๐‘ž,๐‘  ๐ธ | =0 If ๐›ฟ(๐‘ž)โˆˆ๐‘„, then every location (๐‘ž, ๐‘ )โˆˆ [๐‘‰] has exactly one successor: ๐‘ž,๐‘  โˆˆ[๐‘‰] | ๐‘ž,๐‘  ๐ธ | โ‰ค ๐‘† โ€ฒ โ‰ค ๐‘… +1 If ๐›ฟ(๐‘ž)= ๐‘ž 1 โˆง ๐‘ž 2 ๐‘œ๐‘Ÿ ๐›ฟ(๐‘ž)= ๐‘ž 1 โˆจ ๐‘ž 2 for some ๐‘ž 1 , ๐‘ž 2 โˆˆ๐‘„, then we have: ๐‘ž,๐‘  โˆˆ[๐‘‰] | ๐‘ž,๐‘  ๐ธ | โ‰ค2 ๐‘† โ€ฒ โ‰ค2 ๐‘… +1 If ๐›ฟ ๐‘ž =โˆŽ ๐‘ž โ€ฒ ๐‘œ๐‘Ÿ โˆ†๐‘žโ€ฒ, then โˆ€ ๐‘ž,๐‘  โˆˆ ๐‘‰ , ๐‘ž,๐‘  ๐ธ = ๐‘ž โ€ฒ ร—๐‘ ๐‘…= ๐‘ ๐‘… : ๐‘ž,๐‘  โˆˆ[๐‘‰] | ๐‘ž,๐‘  ๐ธ | โ‰ค ๐‘ โˆˆ๐‘† ๐‘ ๐‘… = ๐‘… To sum up: ๐‘ž,๐‘  โˆˆ[๐‘‰] ๐‘ž,๐‘  ๐ธ โ‰ค 2 ๐‘… +1 , ๐ธ โ‰ค2|๐‘„|( ๐‘… +1)

27 The Word problem There is an algorithm which computes in time: ๐’ช ๐‘‘ ๐‘„ ๐‘… ๐‘„ ๐‘† +1 ๐‘‘ ๐‘‘ 2 and in space ๐’ช ๐‘‘ ๐‘„ ๐‘† log ๐‘„ ๐‘† ๐ธ โ‰ค2|๐‘„|( ๐‘… +1) Jurdziยดnskiโ€™s algorithm: For a parity game with ๐‘› vertices, ๐‘š edges and ๐‘‘โ‰ฅ2 ๐‘๐‘Ÿ๐‘–๐‘œ๐‘Ÿ๐‘–๐‘ก๐‘–๐‘’๐‘  can be solved in time ๐’ช ๐‘‘๐‘š ๐‘› ๐‘‘ ๐‘‘ and space: ๐’ช ๐‘‘๐‘› log ๐‘› And now its easy to apply: ๐‘›= ๐‘„ ๐‘† ;๐‘š=2 ๐‘„|( ๐‘… +1)

28 Complementation problem
Theorem: Let ๐’œ=(๐‘„, ๐‘ž ๐ผ , ๐›ฟ,ฮฉ) be an alternating tree automaton. There is an alternating tree automaton ๐’œ =(๐‘„, ๐‘ž ๐ผ , ๐›ฟ , ฮฉ ) such that ๐’œ accepts the complement of the language of ๐’œ. Proof: Lets start with priority function ฮฉ : โˆ€๐‘žโˆˆQ ฮฉ ๐‘ž =ฮฉ+1 The transition function ๐›ฟ: โˆ† โˆŽ โˆŽ โˆ†

29 Complementation problem
Theorem: Let ๐’œ=(๐‘„, ๐‘ž ๐ผ , ๐›ฟ,ฮฉ) be an alternating tree automaton. There is an alternating tree automaton ๐’œ =(๐‘„, ๐‘ž ๐ผ , ๐›ฟ , ฮฉ ) such that ๐’œ accepts the complement of the language of ๐’œ. Let ๐’ฎ be a pointed transition system and ๐’ข=( ๐‘‰ 0 , ๐‘‰ 1 ,๐ธ , ฮฉ, ๐‘ž ๐ผ , ๐‘  ๐ผ ) be the parity game which determines whether A accepts ๐’ฎ. We show that ๐’œ accepts ๐’ฎ iff ๐’œ does not accept ๐’ฎ. We examine the parity game ๐’ข which determines whether ๐’œ accepts ๐’ฎ. Intuitively, we simply change the ownership of every location, and we increase every priority by 1. Let ๐‘‰= ๐‘‰ 0 โˆช ๐‘‰ 1 be the locations of ๐’ข and ๐’ข . Let ๐‘‰ โ€ฒ โŠ†๐‘‰ be the locations ๐‘ฃ(๐‘ž, ๐‘ )โˆˆ๐‘‰ with ๐›ฟ(๐‘ž)=๐‘žโ€ฒ for some ๐‘žโ€ฒโˆˆ๐‘„. We do not change the ownership of locations in Vโ€™. The automaton ๐’œ accepts ๐’ฎ iff there is winning strategy for Player 0 in the parity game: ๐’ข =( V 1 โˆช ๐‘‰ โ€ฒ , ๐‘‰ 0 \ V โ€ฒ ,๐ธ , ฮฉ , ๐‘ž ๐ผ , ๐‘  ๐ผ ) Because parity games are determined, we have to show that there is a winning strategy for Player 0 in ๐’ข iff there is a winning strategy for Player 1 in ๐’ข .

30 Complementation problem
Assuming a winning strategy for Player 0 in ๐’ข we will construct a winning strategy for Player 1 in ๐’ข : Let ๐‘“ 0 : ๐‘‰ 0 โ†’๐‘‰ be a winning strategy for Player 0 in ๐’ข. There is a unique extension ๐‘“โ€ฒ 0 : ๐‘‰ 1 โˆช๐‘‰โ€ฒโ†’ ๐‘‰. Now, assume some play ๐œ‹ in ๐’ข which is consistent with ๐‘“โ€ฒ 0 . Then, ๐œ‹ in ๐’ข is consistent with ๐‘“0, and thus ๐œ‹ is won by Player 1. Conversely, let ๐‘“ 1 : ๐‘‰ 1 โ†’๐‘‰ be a winning strategy for Player 1 in ๐’ข . Clearly, the restriction of ๐‘“ 1 ๐‘ก๐‘œ ๐‘‰ 1 \ ๐‘‰โ€ฒ is a winning strategy for Player 0 in ๐’ข. Consequently, Player 0 has a winning strategy in ๐’ข iff player 1 has a winning strategy in ๐’ข . We can also (but we will not) show that alternating tree automatons are closed under intersection and union.

31 The Emptiness Problem On this section:
The Emptiness Problem complexity analysis

32 The Emptiness Problem Lets start by fixing ๐’œ= ๐‘„, ๐‘ž ๐ผ ,๐›ฟ,ฮฉ
Definition of an inflating transition condition: We call a transition condition โ€œinflatingโ€ if its in the form of ๐›ฟ ๐‘ž =๐‘žโˆง๐‘ž ๐‘œ๐‘Ÿ ๐‘žโˆจ๐‘ž Lemma: For every alternating tree automaton ๐’œ=(๐‘„, ๐‘ž ๐ผ , ๐›ฟ,ฮฉ) there is an automaton ๐’œโ€ฒ=(๐‘„, ๐‘ž ๐ผ , ๐›ฟโ€ฒ,ฮฉ) with ๐ฟ(๐’œ)=๐ฟ(๐’œโ€ฒ) such that for every ๐‘žโˆˆ๐‘„ ๐›ฟ(๐‘ž) is not inflated. Proof: very similar to the complement problem: First we will define ๐›ฟ โ€ฒ :๐‘„โ†’๐‘‡ ๐ถ ๐‘„ ๐‘“๐‘œ๐‘Ÿ ๐‘ž

33 The Emptiness Problem Lemma: For every alternating tree automaton ๐’œ=(๐‘„, ๐‘ž ๐ผ , ๐›ฟ,ฮฉ) there is an automaton ๐’œโ€ฒ=(๐‘„, ๐‘ž ๐ผ , ๐›ฟโ€ฒ,ฮฉ) with ๐ฟ(๐’œ)=๐ฟ(๐’œโ€ฒ) such that for every ๐‘žโˆˆ๐‘„ ๐›ฟ(๐‘ž) is not inflated. Let ๐’ฎ be some pointed transition system. We want to show that ๐’œ accepts ๐’ฎ iff ๐’œโ€™ accepts ๐’ฎ. At first, we observe that ๐’œ has the same behavior ๐‘‰ on ๐’ฎ as ๐’œโ€ฒ. Let ๐’ข=( ๐‘‰ 0 , ๐‘‰ 1 ,๐ธ), ฮฉ, ( ๐‘ž ๐ผ , ๐‘  ๐ผ ) be the parity game to determine whether ๐’œ accepts ๐’ฎ. Let ๐‘‰ โ€ฒ ={๐‘ฃ(๐‘ž, ๐‘ )โˆˆ ๐‘‰ 1 | ๐›ฟ(๐‘ž)=๐‘žโ€ฒโˆง๐‘žโ€ฒ ๐‘“๐‘œ๐‘Ÿ ๐‘ ๐‘œ๐‘š๐‘’ ๐‘žโ€ฒโˆˆ๐‘„}. The locations in ๐‘‰โ€ฒ have exactly one successor. The parity game ๐’ข โ€ฒ= V 0 โˆช ๐‘‰ โ€ฒ , ๐‘‰ 1 \ V โ€ฒ ,๐ธ ,ฮฉ, ๐‘ž ๐ผ , ๐‘  ๐ผ determines whether ๐’œโ€ฒ accepts ๐’ฎ. From now and on its exactly the same proof as the Complementation problem

34 The Emptiness Problem Lets start by fixing ๐’œ= ๐‘„, ๐‘ž ๐ผ ,๐›ฟ,ฮฉ
Definition of an inflating transition condition: We call a transition condition โ€œinflatingโ€ if its in the form of ๐›ฟ ๐‘ž =๐‘žโˆง๐‘ž ๐‘œ๐‘Ÿ ๐‘žโˆจ๐‘ž According to the lemma we can assume that our ๐’œ has no inflate transition conditions. Another notation: Tiles A tile over ๐‘„ is a graph ๐œ—=( ๐‘‰ ๐œ— , ๐ธ ๐œ— ) where ๐‘‰ ๐œ— โŠ†๐‘„, ๐ธโŠ† ๐‘‰ ๐œ— ร— ๐‘‰ ๐œ— and: A tile with port is a tuple (๐œ—,๐‘ž) s.t ๐œ—= ๐ธ ๐œ— , ๐‘‰ ๐œ— ๐‘Ž๐‘›๐‘‘ ๐‘žโˆˆ ๐‘‰ ๐œ— โˆฉ ๐‘„ ฮ”

35 The Emptiness Problem A tile with port is a tuple (๐œ—,๐‘ž) s.t ๐œ—= ๐ธ ๐œ— , ๐‘‰ ๐œ— ๐‘Ž๐‘›๐‘‘ ๐‘žโˆˆ ๐‘‰ ๐œ— โˆฉ ๐‘„ ฮ” ๐‘„ ฮ” โ‰œ{๐‘žโˆˆQ|โˆƒ ๐‘ž โ€ฒ โˆˆ๐‘„:๐›ฟ ๐‘ž =ฮ”๐‘žโ€ฒ } ๐‘„ โˆŽ โ‰œ{๐‘žโˆˆQ|โˆƒ ๐‘ž โ€ฒ โˆˆ๐‘„:๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ } We also define ๐‘ž , ๐‘‰ : ๐‘ž โ‰œ ๐‘ž โ€ฒ ๐‘–๐‘“ ๐›ฟ ๐‘ž =ฮ” ๐‘ž โ€ฒ ๐‘œ๐‘Ÿ ๐›ฟ ๐‘ž =โˆŽ๐‘žโ€ฒ For subsets ๐‘‰โŠ†๐‘„ ๐‘ค๐‘’ ๐‘‘๐‘’๐‘“๐‘–๐‘›๐‘’ ๐‘‰ โ‰œ{ ๐‘ž |๐‘žโˆˆV} All tiles will be: ฮ˜, all tiles with port will be: ฮ˜ ๐‘ We will call a tile with port ๐œ— 0 = ๐‘‰ ๐œ—0 , ๐ธ ๐œ—0 , ๐‘ž 0 and a tile ๐œ— 1 =( ๐‘‰ ๐œ—1 , ๐ธ ๐œ—1 ) (it can be a tile with port as well) concatenable iff ๐‘ž 0 โˆˆ ๐‘‰ ๐œ—1 ๐‘Ž๐‘›๐‘‘ ๐‘‰ ๐œ—0 โˆฉ ๐‘„ โˆŽ โŠ† ๐‘‰ ๐œ—1 i.e: the port must be able to make delta\square transition into ๐‘‰ ๐œ—1 state and that the state intersection of ๐‘‰ ๐œ—0 ๐‘Ž๐‘›๐‘‘ ๐‘„ โˆŽ that make delta\square transition must be contained in ๐‘‰ ๐œ—1

36 The Emptiness Problem Let ๐‘”=( ๐œ— 1 , ๐‘ž 1 ), ( ๐œ— 2 , ๐‘ž 2 ), ใƒปใƒปใƒปโˆˆ ฮ˜ ๐œ” be an infinite sequence of tiles with port where ( ๐œ— ๐‘– , ๐‘ž ๐‘– ) ๐‘Ž๐‘›๐‘‘ ( ๐œ— ๐‘–+1 , ๐‘ž ๐‘–+1 ) are concatenable for every ๐‘–โˆˆ๐œ”. We define the graph of ๐’ˆ in a usual way: ๐‘‰โ‰œ ๐‘–โˆˆ๐œ” ๐‘– ร— ๐‘‰ ๐‘– ๐ธโ‰œ We call an infinite path ๐œ‹ in (๐‘‰, ๐ธ) even iff the maximal priority which occurs in ๐œ‹ infinitely often is even. We call the sequence ๐‘” even iff every infinite path ๐œ‹ ๐‘–๐‘› (๐‘‰, ๐ธ) is even.

37 The Emptiness Problem Theorem: There is a deterministic parity ๐œ”โˆ’๐‘Ž๐‘ข๐‘ก๐‘œ๐‘š๐‘Ž๐‘ก๐‘œ๐‘› ๐ถ with 2 ๐’ช( ๐‘„ 4 log ๐‘„ ) states and priorities bounded by ๐’ช( ๐‘„ 4 ) which accepts a sequence of concatenable tiles ๐‘”โˆˆ ฮ˜ ๐œ” iff ๐‘” is even. Proof concept: construct non-deterministic parity ฯ‰-automaton B โ†’ construct C by a determinization and a complementation of B B states are ๐‘„ร— 0,โ€ฆ, ๐‘„ . (m= ๐‘„ |๐‘„+1|) We specify the transition function ๐›ฟ by a set of triples. Let ( ๐‘ž 1 , ๐‘– 1 ), ๐‘ž 2 , ๐‘– 2 be two states of B, and let (๐‘‰, ๐ธ, ๐‘ž) be a tile with port. There is a transition ๐‘ž 1 , ๐‘– 1 , ๐‘‰, ๐ธ, ๐‘ž , ๐‘ž 2 , ๐‘– 2 in B iff:

38 The Emptiness Problem non-deterministic parity ฯ‰-automaton
Where ๐‘š= ๐‘„ 2 we get: 2 ๐’ช( ๐‘„ 4 log ๐‘„ ) states and priorities bounded by ๐’ช( ๐‘„ 4 )

39 done


Download ppt "Alternating tree Automata and Parity games"

Similar presentations


Ads by Google