Replaying Type Sequences in Forward Heuristic Planning

Tomas de la Rosa, Daniel Borrajo, Angel Garcia Olaya

Heuristic Planning is nowadays one of the top approaches for AI Planning. Although current heuristic planners are quite efficient, the time spent in computing heuristics is still an issue, since this task must be repetitively done for each state explored in the search process. We propose that domain type sequences can be learned to support the heuristic search and to avoid continuously computing heuristics. We present in the paper a CBR approach for extracting, retrieving and replaying type sequences within a heuristic search. We also present the results of testing this idea within two of the classical IPC domains.

Subjects: 1.11 Planning; 3.1 Case-Based Reasoning

Submitted: May 17, 2006

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.