Rational Verification with Quantitative Probabilistic Goals

David Hyland, Julian Gutierrez, Shankaranarayanan Krishna, Michael J. Wooldridge

Abstract: We study the rational verification problem for multi-agent systems in a setting where agents have quantitative probabilistic goals. We use concurrent stochastic games to model multi-agent systems and assume players desire to maximise the probability of satisfying their goals, specified using Linear Temporal Logic (LTL). The main decision problem in this setting is whether a given LTL formula is almost surely satisfied on some pure Nash equilibrium of a given game. We prove that this problem is undecidable in the general case, and then characterise the complexity of this problem under various restrictions on strategies. We also study the problem of deciding whether a given strategy profile is a Nash equilibrium in a given game and show that, unlike the previous verification problem, this question is decidable for several common strategy models.