Title: Reasoning about Sensing Actions in Domains with Multi-Valued Fluents
Authors: Tran Cao Son and Phan Huy Tu (NMSU), Xin Zhang (Arizona State University)

Presenter: Tu Phan

Abstract:

Sensing (or knowledge producing) actions have been the topic of intensive
research in reasoning about action and change and planning. Most of current
approaches based on high-level description languages to reasoning actions
concentrate on providing the solution to the frame problem for domains with
Boolean fluents and sensing actions. Although the simplication to domains
with Boolean fluents does not limit the expressiveness of the fluents, this
solution is too cumbersome and not intuitive.

In this paper, we discuss the weakness of current action languages for
sensing actions with respect to modeling domains with multi-valued fluents.
To address this problem, we propose a language with sensing actions and
multi-valued fluents, called AM_k, and provide a transition function based
semantics for AM_k. We define the entailment relationship between action
theories and queries in AM_k, denoted by AM_k. We discuss some complexity
result about the planning problem in AM_k. Finally, we demonstrate the use
of AM_k through examples from the literature.