Controlled Search over Compact State Representations, in Nondeterministic Planning Domains and Beyond

Two of the most efficient planners for planning in nondeterministic domains are MBP and ND-SHOP2. MBP achieves its efficiency by using Binary Decision Diagrams (BDDs) to represent sets of states that share some common properties, so it can plan for all of these states simultaneously. ND-SHOP2 achieves its efficiency by using HTN task decomposition to focus the search. In some environments, ND-SHOP2 runs exponentially faster than MBP, and in others the reverse is true. In this paper, we discuss the following:

• Type: Conference
• Year: 2006
• Where: AAAI
• Authors: Ugur Kuter, Dana S. Nau
• Keywords: AAAI 2006, Binary Decision Diagrams, Intelligent Agents, Nondeterministic Domains, Task Decompositions