Computing Plan-Length Bounds Using Lengths of Longest Paths

Authors

  • Mohammad Abdulaziz Technische Universität München
  • Dominik Berger Technische Universität München

DOI:

https://doi.org/10.1609/aaai.v35i13.17392

Keywords:

Deterministic Planning, Satisfiability Modulo Theories, Other Foundations of Planning, Routing & Scheduling, Satisfiability

Abstract

We devise a method to exactly compute the length of the longest simple path in factored state spaces, like state spaces encountered in classical planning. Although the complexity of this problem is NEXP-Hard, we show that our method can be used to compute practically useful upper-bounds on lengths of plans. We show that the computed upper-bounds are significantly better than bounds produced by state-of-the-art bounding techniques and that they can be used to improve the SAT-based planning.

Downloads

Published

2021-05-18

How to Cite

Abdulaziz, M., & Berger, D. (2021). Computing Plan-Length Bounds Using Lengths of Longest Paths. Proceedings of the AAAI Conference on Artificial Intelligence, 35(13), 11709-11717. https://doi.org/10.1609/aaai.v35i13.17392

Issue

Section

AAAI Technical Track on Planning, Routing, and Scheduling