On a repeated game with state dependent signalling matrices (Q1072461)

The author considers the case of repeated games with lack of information on both sides, generally not being solved up to now. It is assumed in the paper that there are four possible states of nature described as elements of a product set \(\{0,1\}^ 2\), and a product probability \(\pi\) on it. According to that, a state (\(\mu\),\(\nu)\) is chosen at stage 0, and players I and II are informed about the \(\mu\) and \(\nu\), respectively. Further, according to (\(\mu\),\(\nu)\) there are four \(2\times 2\) payoff matrices \({\mathcal A}(\mu,\nu)\) and four signalling matrices \({\mathcal M}(\mu,\nu)\) (common to the players), all depended on the state. The information the players get after each stage n, \(n=1,2,..\). is the (i,j)- th element \(h_{ij}(\mu,\nu)\) of the matrix \({\mathcal M}(\mu,\nu)\), where i and j are moves players I and II at stage n, respectively, and (\(\mu\),\(\nu)\) is the state of nature choosen at stage 0. As a payoff function the expectation of the Cesaro mean of the payoffs on each stage is defined. The main result of the paper is: for any \(\pi\) the maxmin and the minmax exist in the game. New in the model is the consideration a repeated game with non-symmetric information structure and with signalling matrices depended on the state of nature.