AAAI Publications, Twenty-Second International Joint Conference on Artificial Intelligence

Font Size: 
Generalized Planning: Synthesizing Plans that Work for Multiple Environments
Yuxiao Hu, Giuseppe De Giacomo

Last modified: 2011-06-28

Abstract


We give a formal definition of generalized planning that is independent of any representation formalism. We assume that our generalized plans must work on a set of deterministic environments, which are essentially unrelated to each other. We prove that generalized planning for a finite set of environments is always decidable and EXPSPACE-complete. Our proof is constructive and gives us a sound, complete and complexity-wise optimal technique. We also consider infinite sets of environments, and show that generalized planning for the infinite "one-dimensional problems," known in the literature to be recursively enumerable when restricted to finite-state plans, is EXPSPACE-decidable without sequence functions, and solvable by generalized planning for finite sets.

Full Text: PDF