Informatics Report Series


Report   

EDI-INF-RR-1153


Related Pages

Report (by Number) Index
Report (by Date) Index
Author Index
Institute Index

Home
Title:Logics for Action
Authors: Michael Fourman
Date:Dec 2007
Publication Title:Proceedings of the 3rd Indian International Conference on Artificial Intelligence {IICAI}
Publication Type:Conference Paper Publication Status:Other
Page Nos:1223-1237
Abstract:
Logics of action, for reasoning about the effects of state change, and logics of belief, accounting for belief revision and update, have much in common. Furthermore, we may undertake an action because we hold a particular belief, and revise our beliefs in the light of observed consequences of an action. So studies of these two aspects are inevitably intertwined. However, we argue, a clear separation of the two is helpful in understanding their interactions. We give a semantic presentation of such a separation, introducing a semantic setting that supports one logic for describing the effects of actions, which are modeled as changing the values of particular atomic properties, or fluents, and another for expressing more complex facts or beliefs about the world. We use a simple state-logic, to account for state change, and show how it can be integrated with a variety of domain-logics, of fact or belief, for reasoning about the world. State- and domain-logics are linked, syntactically and semantically; but separate. The state-logic, our logic for action, is quantified propositional logic. Bounded existential propositional quantification is used to specify which literals may be established by a given action. This provides a logically transparent account of the treatment of updates introduced by Herzig et al. (2001), which itself provided a treatment of the frame problem extending the uses in event calculi of the forget operator, occlude and release predicates, all published in 1994. Our, purely classical, logical setting can be seen as a recasting of the state-transition model of action, underlying STRIPS and ADL. Our treatment, like ADL, caters for actions with conditional effects, which can handle ramifications unrepresentable in classic STRIPS; it also includes concurrent actions, non-deterministic effects, and domain axioms or integrity constraints, which are not present in ADL.We introduce an operator, novel in this context, that computes fixpoint
Copyright:
IICAI www.iiconference.org this UoE portal version published by permission.
Links To Paper
No links available
Bibtex format
@InProceedings{EDI-INF-RR-1153,
author = { Michael Fourman },
title = {Logics for Action},
book title = {Proceedings of the 3rd Indian International Conference on Artificial Intelligence {IICAI}},
year = 2007,
month = {Dec},
pages = {1223-1237},
}


Home : Publications : Report 

Please mail <reports@inf.ed.ac.uk> with any changes or corrections.
Unless explicitly stated otherwise, all material is copyright The University of Edinburgh