Improvements to CPCES

Authors

  • Xiaodi Zhang Australian National University
  • Alban Grastien Australian National University

DOI:

https://doi.org/10.1609/socs.v16i1.27289

Keywords:

Problem Solving Using Search

Abstract

This paper introduces three improvements to the conformant planner CPCES, which continuously searches candidate plans and counter-examples against the current candidate plan until a valid plan (no counter-example exists) is found. First, we identify and merge equivalent PDDL facts to accelerate candidate plan generation. Second, we warm-start CPCES by generating multiple carefully selected counter-examples at the beginning of the procedure, which reduces the number of calls to the classical planner. Third, we investigate the use Fast Downward (FD) as the candidate plan generator; in particular, we propose an incremental procedure to generate the SAS+ file used by FD. Our experimental results show significant improvements for each technique.

Downloads

Published

2023-07-02