Using abstraction for generalized planning

Siddharth Srivastava, Neil Immerman, Shlomo Zilberstein

Research output: Contribution to conferencePaperpeer-review

1 Scopus citations


Given the complexity of planning, it is often beneficial to create plans that work for a wide class of problems. This facilitates reuse of existing plans for different instances of the same problem or even for other problems that are somehow similar. We present novel approaches for finding such plans through search and for learning them from examples. We use state representation and abstraction techniques originally developed for static analysis of programs. The generalized plans that we compute include loops and work for classes of problems having varying numbers of objects that must be manipulated to reach the goal. Our algorithm for learning generalized plans takes as its input an example plan for a certain problem instance and a finite 3-valued first-order structure representing a set of initial states from different problem instances. It learns a generalized plan along with a classification of the problem instances where it works. The algorithm for finding plans takes as input a similar 3-valued structure and a goal test. Its output is a set of generalized plans and conditions describing the problem instances for which they work.

Original languageEnglish (US)
StatePublished - 2008
Externally publishedYes
Event10th International Symposium on Artificial Intelligence and Mathematics, ISAIM 2008 - Fort Lauderdale, FL, United States
Duration: Jan 2 2008Jan 4 2008


Other10th International Symposium on Artificial Intelligence and Mathematics, ISAIM 2008
Country/TerritoryUnited States
CityFort Lauderdale, FL

ASJC Scopus subject areas

  • Artificial Intelligence
  • Applied Mathematics


Dive into the research topics of 'Using abstraction for generalized planning'. Together they form a unique fingerprint.

Cite this