Abstractions, Decisions, and Uncertainty
Papers from the AAAI Workshop
Christopher Geib, Chair
Technical Report WS-97-08
130 pp., $30.00
ISBN 978-1-57735-035-4
[Add to Cart] [View Cart]
Classical AI planning, decision-theoretic planning, and even classical decision theory have recognized the need for the use of abstraction in solving large problems. Abstraction in these areas has taken a number of different forms: state space abstraction, construction of partial policies, partial evaluation of policies, action abstraction, action hierarchies, and plan space abstraction. Given the need for abstraction in large problem spaces and the recent advances in decision-theoretic planning, this workshop focuses on abstraction methods for decision making under uncertainty. The goal of this workshop is therefore to bring together researchers working on differing kinds of abstraction for decision making under uncertainty in order to understand what has already been accomplished, outline current problems, and identify promising future directions.
Issues of interest to this workshop
- the use of problem structure to enable abstraction
- state space aggregation and abstraction
- construction and use of abstract policies
- construction and use of abstract actions
- construction and use of action hierarchies
- construction and use of abstract value/utility functions
- plan space abstraction
- uses of abstraction for efficiency, explanation, anytime reasoning,
- ability to handle incomplete information