Three-Valued Abstractions of Games: Uncertainty, but with Precision

  • Authors:
  • Luca de Alfaro;Patrice Godefroid;Radha Jagadeesan

  • Affiliations:
  • University of California, Santa Cruz;Bell Laboratories, Lucent Technologies;School of CTI, DePaul University

  • Venue:
  • LICS '04 Proceedings of the 19th Annual IEEE Symposium on Logic in Computer Science
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present a framework for abstracting two-player turn-based games that preserves any formula of the alternating 驴-calculus (AMC). Unlike traditional conservativ abstractions which can only prove the existence of winning strategies for only one of the players, our framework is based on 3-valued games, and it can be used to prove and disprove formulas of AMC including arbitrarily nested strategy quantifiers. Our main contributions are as follows.We define abstract 3-valued games and an alternating refinement relation on these that preserves winning strategies fo both players.We provide a logical characterization of the alternating refinement relation. We show that our abstractions are as precise as can be via completeness results.We present AMC formulas that solve 3-valued games with w-regular objectives, and we show that such games are determined in a 3-valued sense.We also discuss the complexity of model checking arbitrary AMC formulas on 3-valued games and of checking alternating refinement.