Year: 2026
Type: preprint
Abstract: This paper studies a two-person zero-sum stochastic game, which must terminate when either player quits the game or a state of the game first enters a prescribed target set. The performance criterion is the probability that the total reward reaches a given threshold. First, under a mild condition, we prove the existence of the value and a Nash equilibrium of the game by developing a comparison theorem. Second, for solving the game with infinite states, we develop a finite-state approximation method based on truncation, and derive explicit error bounds between the values of the truncated and original games. Third, using the error bounds, we design an effective algorithm, by which we can not only obtain any approximation of the value and Nash equilibria of the original game, but also provide convergence analysis of the approximation. Finally, a queueing example is used to illustrate the advantage of our condition and the effectiveness of our algorithm.
Source: arXiv (Cornell University)
Authors: Wen, Xin, Guo, Xianping, Guo, Xin, Zheng, Zewu
Institutions: Sun Yat-sen University, Chinese University of Hong Kong
Language: English
Topic: Game Theory and Applications
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7235533595
