scientific article; zbMATH DE number 1329117

Publication date 3 April 2001

\(N\)-person non-cooperative games under uncertainty Hurwitz pessimism-optimism principle maximin principle Nash solution Nash-Hurwitz equilibrium Slater-maximal two person linear quadratic game

Mathematics Subject Classification ID

(n)-person games, (n>2) (91A06) Noncooperative games (91A10)

Review text

In the context of \(N\)-person non-cooperative games under uncertainty, the paper formalizes a solution concept which incorporates the features of the Nash solution and the Hurwitz pessimism-optimism principle. Moreover, this is extended to incorporate a vector analogue of the maximin principle. Sufficient conditions for the existence of such a solution, in pure strategies, is provided. More formally, let \(\langle N, \{X_i\}_{i\in N},Y, \{f_i(x,y)\}_{i\in N}\rangle\) denote an \(N\)-person non-cooperative game under uncertainty, where \({\mathbf N}= \{1,\dots,N\}\) is the index set of players, \(X_i\subset \mathbb{R}^{n_i}\) is the set of strategies \(x_i\) of player \(i\), \(X\equiv \prod_{i=1}^N X_i\) is the set of situations \(x\equiv (x_1,\dots, x_N)\) of the game, \(Y\subset \mathbb{R}^m\) is the set of uncertainties \(y\); and \(f_i(x,y): X\times Y\to \mathbb{R}\) is the payoff function of player \(i\). For each \(i= 1,\dots, N\), let \(\alpha_i\) denote the pessimism-optimism index of player \(i\) and define the function \(g_i:X\to \mathbb{R}\) by \[ g_i(x)\equiv \alpha_i \min_{y\in Y} f_i(x,y)+ (1-\alpha_i) \max_{y\in Y} f_i(x,y), \quad\text{for }x\in X. \] Given a situation \(x^0\in X\) and a strategy \(x_i\in X_i\), \((x^0\parallel x_i)\) denotes the situation where \(x_i^0\) is replaced by \(x_i\), i.e. \((x^0\parallel x_i)\equiv (x_1^0,\dots, x_{i-1}^0, x_i, x_{i+1}^0,\dots, x_N^0)\). A Nash-Hurwitz (N-H) equilibrium is a situation \(x^0\in X\) satisfying \(g_i (x^0\parallel x_i)\leq g_i(x^0)\) for any \(x_i\in X_i\), \(i\in N\); it is Slater-maximal if, for any N-H equilibrium \(x\), we have \(g_i(x^0)\geq g_i(x)\) for some \(i\in N\). Introducing probability measures on the strategy spaces \(X_i\) these concepts are extended in the usual way over the game when mixed strategies are considered. It is shown that: Theorem: When the sets \(X_i\) \((i\in N)\), \(Y\) are compact and the payoff functions \(f_i(x,y)\) are continuous on \(X\times Y\), a Slater-maximal N-H equilibrium exists in mixed strategies. Define the auxiliary functions \(h_i: X\times Y\times Y\to \mathbb{R}\) as follows: \(h_i: (x,y^1,y^2)= \alpha_if_i (x,y^1)+\) \((1-\alpha_i) f_i(x,y^2)\). A situation \(x^*\in X\) is called a guaranteed N-H equilibrium if there exist vectors \((y_1^1,\dots, y_N^1)\) and \((y_1^2,\dots, y_N^2)\) in \(Y^N\) such that: \(h_i(x^*\parallel x_i, y_i^1,y)\leq h_i (x^*, y_i^1, y_i^2)\) for any \(x_i\in X_i\), \(y\in Y\), \(i\in N\) and \(f_i(x^*,y_i^1)\leq f_i(x^*,y)\) for any \(y\in Y\), \(i\in N\). Any guaranteed N-H equilibrium is shown to be N-H equilibrium. Furthermore, it is established that: Proposition: If the sets \(X_i\), \(i\in N\), and \(Y\) are convex and compact, the functions \(f_i(\cdot)\) are continuous on \(X\times Y\), \(f_i(\cdot)\) is concave with respect to \((x_i,y)\) given \(x_j\) for \(j\neq i\); and \(f_i(x,y)\) are linear with respect to \(y\) for every \(x\in X\), then there exists then there exists a guaranteed N-H equilibrium situation in pure strategies. A two person linear quadratic game, in which the payoff functions have quadratic and linear terms in strategies and uncertain variables, is examined in detail and explicit solutions for the guaranteed N-H equilibrium are obtained. See also the journal version published in Int. J. Game Theory Algebra 10, No. 2, 89-100 (2000) above (Zbl 0958.91004). For the entire collection see [Zbl 0915.00030].

Recommendations

Cited in

(11)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4257482)