A Method for Solving Nash Equilibrium of the Game Oriented to Multi-agent Interactions

Jin LI,Kun YUE,Wei-Yi LIU
DOI: https://doi.org/10.3969/j.issn.1002-137X.2007.03.049
2007-01-01
Computer Science
Abstract:Among the existing methods for computing the Nash Equilibrium of graphical games, agents play discretized mixed Strategies. Consequently, only an approximate Nash Equilibrium can be achieved. In this paper, we induce the problem of obtaining an exact Nash Equilibrium into a generic function optimization in a space of continuous strategies. Meanwhile, the objective is defined as the summation of utility regret in strategic profiles, and its corresponding optimized solutions are achieved as Nash equilibrium. Thus, the gradient descent of objective function can be calculated through a set of linear program. Further, a gradient descent algorithm to find exact Nash Equilibrium of Graphical Game with any structure is presented. Our proposed method can be used to get the exact Nash equilibrium of graphical games with arbitrary structures. As well, our method has fairly good precision and efficiency even on the large-scaled graphical games.
What problem does this paper attempt to address?