Download presentation
Presentation is loading. Please wait.
Published byMadyson Laws Modified over 9 years ago
1
Social Behavior of Agents and Stable Models Francesco Buccafurri and Gianluca Caminiti DIMET, Università degli Studi “Mediterranea” di Reggio Calabria Convegno Italiano di Logica Computazionale (CILC’05) Roma, 21-22 Giugno 2005
2
2CILC'05, Roma, 21-22 Giugno 2005 MAS & Logic Programming Agents Logic Programs Agents Logic Programs Desires/Requests Fixpoints The behavior of one agent can depend on that of the other agents. Social Ability: interaction enables reasoning on agents’ mental states.
3
3CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/0 a1a1a1a1 A |A| = N
4
4CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/1 head ← [l,h]{body} a1a1a1a1 A
5
5CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/3 head ← [l,h]{body} a1a1a1a1 S l |S| h A
6
6CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/4 head ← [l,h]{body} a1a1a1a1 S l |S| h A
7
7CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/4 head ← [l,h]{body} a1a1a1a1 S l |S| h A
8
8CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/5 head ← [a 2 ]{body} a1a1a1a1 A a2a2a2a2
9
9CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/5 a1a1a1a1 A a2a2a2a2 head ← [a 2 ]{body}
10
10CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/5 a1a1a1a1 A a2a2a2a2 head ← [a 2 ]{body}
11
11CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/6 head ← [l 1,h 1 ]{body 1, [l 2,h 2 ]{body 2 }} a1a1a1a1 A
12
12CILC'05, Roma, 21-22 Giugno 2005 Social-Oriented Reasoning/6 a1a1a1a1 S1S1S1S1 l 1 |S 1 | h 1 l 2 |S 2 | h 2 S 2 S 1 A S2S2S2S2 head ← [l 1,h 1 ]{body 1, [l 2,h 2 ]{body 2 }}
13
13CILC'05, Roma, 21-22 Giugno 2005 Example 1: a Wedding Party P 1 (Agent 1 ) party ← [ N/2 - 1, ]{party} P 2 (Agent 2 ) okay(party) ← okay(drive) ← party P 3 (Agent 3 ) party ← [Agent 2 ]{party, not drive} P 4 (Agent 4 ) empty program
14
14CILC'05, Roma, 21-22 Giugno 2005 Example 1: Intended Models {},{}, {party P 1, party P 2, drive P 2 },{party P 1, party P 2, drive P 2 }, {party P 1, party P 2, party P 3 }.{party P 1, party P 2, party P 3 }.
15
15CILC'05, Roma, 21-22 Giugno 2005 Example 2: a P2P Scenario download(X) ← [min, ]{ share(X), [1,]{not incomplete(X)} }, file(X) okay(share(X)) ← [0.33*N, ]{ share(X), [0.1*N,0.2*N]{high_bw} }, file(X)
16
16CILC'05, Roma, 21-22 Giugno 2005 Syntax: Social Rules h ← body body = b 1, …, b m, s 1, …, s k (m ≥ 0, k ≥0) body = b 1, …, b m, s 1, …, s k (m ≥ 0, k ≥ 0) h - literal or okay(p) h - literal or okay(p) b i (1 ≤ i ≤ m) - (possibly NAF) literal b i (1 ≤ i ≤ m) - (possibly NAF) literal s j (1≤ j ≤ k) - (possibly NAF) SSC s j (1 ≤ j ≤ k) - (possibly NAF) SSC SOLP program = set of social rules SOLP collection = set of SOLP programs
17
17CILC'05, Roma, 21-22 Giugno 2005 Syntax: SSC [selection_condition]{body} [selection_condition] [l, h] [agent_id] Warning:[1, 1] ≠ [id]. Note: SSCs can be nested.
18
18CILC'05, Roma, 21-22 Giugno 2005 Semantics: Autonomy P - SOLP program Var(P) - atoms occurring in P AP - autonomous reduction of P AT P - extends classical T P to social rules AFP(P) = set of autonomous fixpoints of P
19
19CILC'05, Roma, 21-22 Giugno 2005 Semantics: SOLP Collection C ={P 1,..., P n } - a SOLP collection Social Interpretation for C - I = I P 1 … I P n I P j - a labeled interpretation for P j C (1 ≤ j ≤ n) Candidate Social Interpretations for C – U(P 1, …, P n ) = {F 1 P 1 … F n P n | F i AFP(P i ), 1 ≤ i ≤ n}
20
20CILC'05, Roma, 21-22 Giugno 2005 Semantics: literals C ={P 1,..., P n } - a SOLP collection I - a social interpretation for C a Var(P i ), P i C (resp. not a) is true w.r.t. I if a P i I (resp. a P i I) if a P i I (resp. a P i I)
21
21CILC'05, Roma, 21-22 Giugno 2005 Semantics: SSCs [selection_condition]{body} is true w.r.t. I if a set of agents exists s.t. 1) selection_condition holds, and 2) body is true w.r.t. I C ={P 1,..., P n } - a SOLP collection I - a social interpretation for C
22
22CILC'05, Roma, 21-22 Giugno 2005 Semantics: Social Rules C ={P 1,..., P n } - a SOLP collection I - a social interpretation for C h ← body - a social rule r of P C r is true w.r.t. I if either: h is true w.r.t. I, or h is true w.r.t. I, or body is false w.r.t. I. body is false w.r.t. I.
23
23CILC'05, Roma, 21-22 Giugno 2005 Semantics: Social Models I U(P 1, …, P n ) is a social model of C if r P 1 … P n, r is true w.r.t. I C ={P 1,..., P n } - a SOLP collection I - a social interpretation for C
24
24CILC'05, Roma, 21-22 Giugno 2005 Translation: Overview Goal: to build a single logic program J(C) whose stable models are in 1:1 correspondence with the social models of C. C ={P 1,..., P n } - a SOLP collection
25
25CILC'05, Roma, 21-22 Giugno 2005 Translation: Step-by-step Each SOLP program is rewritten as a classical one, P’, where the SSCs are represented by special literals. Each SOLP program is rewritten as a classical one, P’, where the SSCs are represented by special literals. Conventions on atom names are used in order to know which program a given atom comes from. Conventions on atom names are used in order to know which program a given atom comes from. All the SSCs in C are rewritten as sequences of DLP A aggregate functions in a logic program S. All the SSCs in C are rewritten as sequences of DLP A aggregate functions in a logic program S.
26
26CILC'05, Roma, 21-22 Giugno 2005 Translation: Final Result J(C) = ( P C P’) S The logic program J(C) selects, among all the candidate social interpretations for C, those w.r.t. which all the SSCs in C are true. Feasible in polynomial time
27
27CILC'05, Roma, 21-22 Giugno 2005 Translation: Theorem J(C) = ( P C P’) S A one-to-one correspondence exists between the social models in SOS(P 1,..., P n ) and the stable models of J(C). C ={P 1,..., P n } - a SOLP collection SOS(P 1,..., P n ) - the social models of C
28
28CILC'05, Roma, 21-22 Giugno 2005 Social Models & Joint Fixpoints COLP* program = logic program + okay rules COLP program desires/consents of a single agent Semantics of a set of COLP programs (Joint Fixpoint Semantics*): common agreement among the agents. * F. Buccafurri and G. Gottlob. Multiagent Compromises, Joint Fixpoints and Stable Models, volume 2407 of LNCS and LNAI. Springer, 2002
29
29CILC'05, Roma, 21-22 Giugno 2005 Social Semantics extends JFP From: COLP programs (JFP Semantics) To:SOLP programs (Social Semantics) Feasible in polynomial time
30
30CILC'05, Roma, 21-22 Giugno 2005 Social Models & JFP: Theorem A one-to-one correspondence exists between the Joint Fixpoints of P 1,..., P n and the social models of SOS(Q 1,..., Q n ). P 1,..., P n - COLP programs Q 1,..., Q n - SOLP programs (the translation of P 1,..., P n )
31
31CILC'05, Roma, 21-22 Giugno 2005 SOS Existence: Theorem Theorem The problem SOS n is NP complete. Problem SOS n Instance: A SOLP collection P 1,..., P n Question: Is SOS(P 1,..., P n ) ?
32
32CILC'05, Roma, 21-22 Giugno 2005 Conclusions SOcial Logic Programming (SOLP) enables social behavior among a community of agents. SOcial Logic Programming (SOLP) enables social behavior among a community of agents. SOLP extends COLP. SOLP extends COLP. Complexity results. Complexity results.
33
33CILC'05, Roma, 21-22 Giugno 2005 Finally… Thank You!
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.