A Framework for Planning with Incrementally Created Graphs in Attributed Problem Spaces

Share
Author
Balakirsky, S.
Pub. date
July 2003
Pages
280
Binding
hardcover
Volume
270 of Dissertations in Artificial Intelligence
ISBN print
978-1-58603-370-5
Subject
Artificial Intelligence, Computer & Communication Sciences, Computer Science

In this publication, a framework for parallel planning agents is developed and applied to planning problems ranging from domain independent planning to planning for autonomous vehicle systems. This general-purpose framework combines both logic-based and cost-based planning approaches in order to allow for the creation of logic-constrained, cost optimal plans with respect to possibly dynamic environments, user objectives, and constraints while providing the ability to construct “any-time” plans in partitioned problem spaces. Experiments are performed in rule-based and cost-based planning domains with examples from the literature such as the towers of Hanoi, and domain independent planning as well as current state-of-the-art real-time autonomous vehicle applications such as on- and off-road driving. These experimental results show that this incremental technique produces smaller planning graphs with higher quality planning results than standard graph-based planning techniques. These improvements are the result of the incremental graph construction approach and the application of hard and soft constraints that are also shown to elicit a variety of different driving behaviors in the autonomous vehicle domain and different problem solving approaches in the towers of Hanoi domain.