Solving Simple Planning Problems with More Inference and No Search

Many problems used in AI planning including Blocks, Logistics, Gripper, Satellite, and others lack the interactions that characterize puzzles and can be solved nonoptimally in low polynomial time. They are indeed easy problems for people, although as with many other problems in AI, not always easy for machines. In this work, we study the type of inferences that are required in a domain-independent planner for solving simple problems such as these in a backtrack-free manner by performing polynomial node operations. For this, we make use of the optimal temporal planner CPT which combines a POCL branching scheme with strong inference mechanisms, and show that a few simple and general additional inference mechanisms suffice to render the search over various domains backtrack free. This is an interesting empirical finding, we believe, that may contribute to the development of more robust automated planners, and to a better understanding of human planning. Significant performance gains i...

• Type: Conference
• Year: 2005
• Where: CP
• Authors: Vincent Vidal, Hector Geffner
• Keywords: CP 2005, Inference Mechanisms, Low Polynomial Time, Problems, Programming Languages