Abstract
This research extends classical game theory by introducing fuzzy payoffs into two-person non-zero-sum bimatrix games. By utilizing fuzzy set theory, the authors define three distinct concepts of minimax equilibrium strategies. The study demonstrates that these complex fuzzy problems can be effectively solved by converting them into systems of standard, "crisp" mathematical inequalities, providing a bridge between uncertain payoffs and traditional optimization techniques.
Introduction
In the foundational models of game theory established by von Neumann and Nash, players make strategic decisions based on precise, deterministic payoff values. However, real-world applications in economics, management, and social sciences often involve significant uncertainty. Outcomes are frequently approximate or "fuzzy," making it difficult to apply traditional equilibrium concepts.
While fuzzy set theory has previously been integrated into game-theoretic frameworks, defining a stable equilibrium—where no player has an incentive to deviate—remains a challenge when rewards are not single numerical values. This paper addresses this gap by formalizing equilibrium conditions for games where payoffs are represented as symmetric triangular fuzzy numbers.
Methodology and Definitions
The authors employ Fuzzy Set Theory and Bimatrix Game Theory to model uncertainty. The approach is grounded in several key mathematical constructs:
1. Symmetric Triangular Fuzzy Numbers
Payoffs are modeled as $(m, h)_T$, where $m$ represents the center value (the most likely outcome) and $h$ represents the deviation or spread. This structure allows for a "best guess" accompanied by a defined margin of error.
2. Fuzzy Max Order
To compare fuzzy outcomes, the researchers introduce binary relations known as "Fuzzy Max Order." Instead of simple scalar comparisons, they define "strict" and "strong" fuzzy orders based on $\alpha$-level sets—intervals where the membership function of the fuzzy number exceeds a specific threshold.
3. Hierarchy of Equilibrium Strategies
The paper proposes three distinct levels of equilibrium to account for different degrees of strategic stability:
- Minimax Equilibrium: The most stringent and robust definition.
- Non-dominated Minimax Equilibrium: A strategy where no alternative strategy is strictly superior across the fuzzy range.
- Weak Non-dominated Minimax Equilibrium: A relaxed version of the non-dominated strategy, providing a broader set of potential solutions.
The "Crisp" Bridge: Key Results
The core contribution of this work is the mathematical proof that fuzzy equilibrium conditions can be reduced to "crisp" (non-fuzzy) equivalents.
Theorem 3.1 establishes the necessary and sufficient conditions for a minimax equilibrium. It proves that a fuzzy equilibrium exists if and only if a specific set of four linear inequalities involving the centers and deviations of the payoff matrices are satisfied. This allows decision-makers to use well-established linear programming and optimization tools even when dealing with uncertain data.
Furthermore, Theorem 3.4 guarantees the existence of at least one non-dominated (and consequently, at least one weak non-dominated) minimax equilibrium strategy for any such fuzzy bimatrix game. This ensures that the model is not just a theoretical curiosity but a practically applicable tool for decision-making under uncertainty.
Insights and Implications
The study reveals that players in a fuzzy bimatrix game are essentially navigating a pair of traditional two-person zero-sum games with crisp payoffs derived from the boundaries of the fuzzy sets. This decomposition simplifies the conceptual complexity of fuzzy games without sacrificing the nuance of the uncertainty involved.
Limitations and Future Directions
While providing a solid foundation, the research is limited by its focus on symmetric triangular fuzzy numbers. In many practical scenarios, uncertainty is asymmetric (e.g., downside risk may far outweigh upside potential). Additionally, the paper focuses on the theoretical existence and conditions of equilibrium rather than the computational algorithms required for large-scale matrices.
Future work may involve extending these results to $n$-person games or exploring non-triangular fuzzy membership functions to better capture diverse types of real-world uncertainty.
Summary updated