Temporal Epistemic Gossip Problems - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2019

Temporal Epistemic Gossip Problems

Résumé

Gossip problems are planning problems where several agents have to share information (`secrets') by means of phone calls between two agents. In epistemic gossip problems the goal can be to achieve higher-order knowledge, i.e., knowledge about other agents' knowledge; to that end, in a call agents communicate not only secrets, but also agents' knowledge of secrets, agents' knowledge about other agents' knowledge about secrets, etc. Temporal epistemic gossip problems moreover impose constraints on the times of calls. These constraints are of two kinds: either they stipulate that a call between two agents must necessarily be made at some time point, or they stipulate that a call can be made within some possible (set of) interval(s). In the non-temporal version, calls between two agents are either always possible or always impossible. We investigate the complexity of the plan existence problem in this general setting. Concerning the upper bound, we prove that it is in NP in the general case, and that it is in P when the problem is non-temporal and the goal is a positive epistemic formula. As for the lower bound, we prove NP-completeness for two fragments: problems with possibly negative goals even in the non-temporal case, and problems with temporal constraints even if the goal is a set of positive atoms.
Fichier principal
Vignette du fichier
cooper_24721.pdf (352.66 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02378391 , version 1 (25-11-2019)

Identifiants

  • HAL Id : hal-02378391 , version 1
  • OATAO : 24721

Citer

Martin Cooper, Andreas Herzig, Frédéric Maris, Julien Vianey. Temporal Epistemic Gossip Problems. 16th European Conference on Multi-Agent Systems (EUMAS 2018), Dec 2018, Bergen, Norway. pp.1-14. ⟨hal-02378391⟩
55 Consultations
71 Téléchargements

Partager

Gmail Facebook X LinkedIn More