Friday, May 21, 2010
12:30 PM
CSB 703
Alfonso Gerevini
U. of Brescia, Italy
Automated Planning in Temporal Domains: Some Recent Advances and Research Topics
Automated planning is a central area of artificial intelligence, involving
the design of languages and computational models for reasoning about
actions, change and time. Many computational models for plan synthesis have
been developed, and modern planning systems efficiently support languages
that are much more expressive than the simple STRIPS language largely
studied in the past. In this talk, I focus on planning in temporal domains
specified by the Planning Domain Description Language (PDDL), which is a
standard reference for the automated planning community. In particular,
I will consider versions 2.2 and 3 of PDDL. After a brief description of
the relevant features of PDDL2.2 (which allow for various types of
temporal constraints, quality metrics and exogenous events), I will
give an overview of a successful system for solving the class of temporal
planning problems that can be expressed in this language. I will then
discuss some challenges posed by PDDL3 constructs allowing for strong
and soft constraints on goals and plan structure.