KR2022Proceedings of the 19th International Conference on Principles of Knowledge Representation and ReasoningProceedings of the 19th International Conference on Principles of Knowledge Representation and Reasoning

Haifa, Israel. July 31–August 5, 2022.

Edited by

ISSN: 2334-1033
ISBN: 978-1-956792-01-0

Sponsored by
Published by

Copyright © 2022 International Joint Conferences on Artificial Intelligence Organization

Verification and Realizability in Finite-Horizon Multiagent Systems

  1. Senthil Rajasekaran(Rice University)
  2. Moshe Y. Vardi(Rice University)


  1. KR and game theory
  2. KR and autonomous agents and multi-agent systems


The problems of verification and realizability are two central themes in the analysis of reactive systems. When multiagent systems are considered, these problems have natural analogues of existence (nonemptiness) of pure-strategy Nash equilibria and verification of pure-strategy Nash equilibria. Recently, this body of work has begun to include finite-horizon temporal goals. With finite-horizon temporal goals, there is a natural hierarchy of goal representation, ranging from deterministic finite automata (DFA), to nondeterministic finite automata (NFA), and to alternating finite automata (AFA), with a worst-case exponential gap between each successive representation. Previous works showed that the realizability problem with DFA goals was PSPACE-complete, while the realizability problem with temporal logic goals is in 2EXPTIME. In this work, we study both the realizability and the verification problems with respect to various goal representations. We first show that the realizability problem with NFA goals is EXPTIME-complete and with AFA goals is 2EXPTIME-complete, thus establishing strict complexity gaps between realizability with respect to DFA, NFA, and AFA goals. We then contrast these complexity gaps with the complexity of the verification problem, where we show that verification with respect to DFAs, NFA, and AFA goals is PSPACE-complete.