Formally Verified SAT-Based AI Planning

Authors

  • Mohammad Abdulaziz King's College London Technische Universität München
  • Friedrich Kurz Technische Universität München

DOI:

https://doi.org/10.1609/aaai.v37i12.26714

Keywords:

General

Abstract

We present an executable formally verified SAT encoding of ground classical AI planning problems. We use the theorem prover Isabelle/HOL to perform the verification. We experimentally test the verified encoding and show that it can be used for reasonably sized standard planning benchmarks. We also use it as a reference to test a state-of-the-art SAT-based planner, showing that it sometimes falsely claims that problems have no solutions of certain lengths.

Downloads

Published

2023-06-26

How to Cite

Abdulaziz, M., & Kurz, F. (2023). Formally Verified SAT-Based AI Planning. Proceedings of the AAAI Conference on Artificial Intelligence, 37(12), 14665-14673. https://doi.org/10.1609/aaai.v37i12.26714

Issue

Section

AAAI Special Track on Safe and Robust AI