On the convergence to and location of attractors of uncertain, dynamic games

Abstract

The delegation of decision making to distributed control agents has been a standard approach to operating large, complex systems. The performance of these multi-agent systems depends as much on the decomposition of the overall operating task as on the interplay between the agents. To that end, game theory has served as a formalism to model multi-agent systems, understand the innerworking of their agents, and analyze issues of convergence (stability) and location of attractors (optimality). This paper delivers simple transformations and algorithms that allow altruistic agents to induce convergence to Nash equilibrium points, and draw these attractors closer to the set of Pareto efficient solutions, of dynamic games whose agents' problems cannot be anticipated. © Springer-Verlag 2004.