Agent Dynamic Logic (ADL) and
Operational semantics for agent programming languages
Wayne Wobcke
School of Computer Science and Engineering
University of New South Wales
9-30am - 11am,
RMIT University Function Room (near Kaleide Theatre)
Building 8, level 2, RMIT,
Swanston Street, Melbourne.
Abstract:
In recent years, there have been increased efforts towards defining
rigorous operational semantics for a range of agent programming
languages. At the same time, there have been increased efforts to
develop logical frameworks for modelling belief, desire and intention
(and related notions) that make closer connections to the workings of
particular architectures, thus aiming to provide some computational
interpretation of these abstract models. However, there remains a
substantial gap between the more abstract logical approaches and the
more computationally oriented operational approaches.
In this paper, we aim to bridge this gap for a class of agents whose
architecture and programming language is based on the PRS system. We
first develop an operational semantics for a simplified 'PRS-like'
agent architecture, abstracting some common features of this class.
We then present a new logical framework called Agent Dynamic Logic
(ADL) that combines BDI Logic and Propositional Dynamic Logic, and
show how ADL can be used to model the 'mental states' of PRS-like
agents (their beliefs, desires and intentions) and the semantics of
their agent programs. The modelling allows a mapping between the
states in the operational description of the PRS-like architecture and
the models in the logical semantics of the agent's mental states, thus
bridging the gap between the two levels of description for this class
of architectures. Finally, we consider the logical properties of
intentions and various rationality postulates for PRS-like agents that
are a consequence of this modelling, in particular Bratman's asymmetry
thesis, the side effect problem for intentions and Rao and Georgeff's
non-transference principles. We show that PRS-like agents do enjoy
many of the accepted logical properties of intention and rationality
postulates as automatic consequences of our approach, though we argue
that PRS-like agents do not have intentions in the sense required by
Bratman's theory.
Biosketch:
Biosketch not yet available.