Bounded-Memory Strategies in Partial-Information Games

被引:0
作者
Bose, Sougata [1 ]
Ibsen-Jensen, Rasmus [1 ]
Totzke, Patrick [1 ]
机构
[1] Univ Liverpool, Liverpool, England
来源
PROCEEDINGS OF THE 39TH ANNUAL ACM/IEEE SYMPOSIUM ON LOGIC IN COMPUTER SCIENCE, LICS 2024 | 2024年
基金
英国工程与自然科学研究理事会;
关键词
Finite-memory strategies; Equilibria; Approximation; Imperfect information games;
D O I
10.1145/3661814.3662096
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
We study the computational complexity of solving stochastic games with mean-payoff objectives. Instead of identifying special classes in which simple strategies are sufficient to play epsilon-optimally, or form epsilon-Nash equilibria, we consider general partial-information multiplayer games and ask what can be achieved with (and against) finite-memory strategies up to a given bound on the memory. We show NP-hardness for approximating zero-sum values, already with respect to memoryless strategies and for 1-player reachability games. On the other hand, we provide upper bounds for solving games of any fixed number of players k. We show that one can decide in polynomial space if, for a given k-player game, epsilon >= 0 and bound b, there exists an epsilon-Nash equilibrium in which all strategies use at most b memory modes. For given epsilon > 0, finding an epsilon-Nash equilibrium with respect to b-bounded strategies can be done in FNPNP. Similarly for 2-player zero-sum games, finding a b-bounded strategy that, against all b-bounded opponent strategies, guarantees an outcome within.. of a given value, can be done in FNPNP. Our constructions apply to parity objectives with minimal simplifications. Our results improve the status quo in several well-known special cases of games. In particular, for 2-player zero-sum concurrent mean-payoff games, one can approximate ordinary zero-sum values (without restricting admissible strategies) in FNPNP.
引用
收藏
页数:14
相关论文
共 40 条
[21]  
Frederiksen SKS, 2013, LECT NOTES COMPUT SC, V8283, P457, DOI 10.1007/978-3-642-45030-3_43
[22]  
Gillette D, 1958, ANN MATH STUD, VIII, P179, DOI 10.1515/9781400882151-011
[23]  
Gimbert H, 2005, LECT NOTES COMPUT SC, V3653, P428, DOI 10.1007/11539452_33
[24]  
Goldreich O, 2006, LECT NOTES COMPUT SC, V3895, P254, DOI 10.1007/11685654_12
[25]   The Big Match with a Clock and a Bit of Memory [J].
Hansen, Kristoffer Arnsfelt ;
Ibsen-Jensen, Rasmus ;
Neymanc, Abraham .
MATHEMATICS OF OPERATIONS RESEARCH, 2023, 48 (01) :419-432
[26]   The Complexity of Solving Reachability Games Using Value and Strategy Iteration [J].
Hansen, Kristoffer Arnsfelt ;
Ibsen-Jensen, Rasmus ;
Miltersen, Peter Bro .
THEORY OF COMPUTING SYSTEMS, 2014, 55 (02) :380-403
[27]  
Hansen KA, 2011, ACM S THEORY COMPUT, P205
[28]   Winning concurrent reachability games requires doubly-exponential patience [J].
Hansen, Kristoffer Arnsfelt ;
Koucky, Michal ;
Miltersen, Peter Bro .
24TH ANNUAL IEEE SYMPOSIUM ON LOGIC IN COMPUTER SCIENCE, PROCEEDINGS, 2009, :332-+
[29]   Mean-payoff games with partial observation [J].
Hunter, Paul ;
Pauly, Arno ;
Perez, Guillermo A. ;
Raskin, Jean-Francois .
THEORETICAL COMPUTER SCIENCE, 2018, 735 :82-110
[30]  
Madani O, 1999, SIXTEENTH NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-99)/ELEVENTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE (IAAI-99), P541