Paper Info
Title
Interleaving Planning and Plan Execution with Incomplete Knowledge in the Event Calculus.
Abstract
In most A.I. planning approaches it is assumed that the planning agent has complete knowledge about its environment. If this is the case the agent acts in two steps: It first plans and then executes its plan. However, humans do usually not behave this simple, as in most real world problems knowledge about the environment is incomplete. To solve real world problems, acting, sensing and planning has to be interleaved in a cyclic manner: Knowledge has to be acquired during plan execution and the plan has to be adopted incrementally to the acquired knowledge. In our work, we employ the Discrete Event Calculus Knowledge Theory (DECKT) and combine it with a Lazy Branching strategy to interleave planning and plan execution for problems with incomplete knowledge: We make optimistic assumptions for unknown contingencies and lazily postpone the planning for a certain contingency until it is resolved. As the Event Calculus accounts for time, the proposed approach allows to combine planning with incomplete knowledge, concurrency and explicit time.
Year
DOI
Venue
2012
10.3233/978-1-61499-096-3-107
Frontiers in Artificial Intelligence and Applications
Keywords
Field
DocType
Planning,Incomplete Knowledge,Action Theory,Event Calculus
Event calculus,Incomplete knowledge,Programming language,Computer science,Interleaving
Conference
Volume
ISSN
Citations 
241
0922-6389
0
PageRank 
References 
Authors
0.34
0
2
Name
Order
Citations
PageRank
Manfred Eppe16311.60
Dominik Dietrich200.68