Title:Dynamic network analysis of a target defense differential game with limited observations

Abstract:In this paper, we study a Target-Attacker-Defender (TAD) differential game involving one attacker and one target, both with unlimited visibility range, and multiple defenders with limited visibility capabilities. We assume that the target and attacker are unaware of defenders' visibility constraints which results in a dynamic game with asymmetric information. We seek to obtain the strategies that are likely to be used by the players. The visibility constraints of the defenders induce a visibility network which encapsulates the visibility information during the evolution of the game. Based on this observation, we introduce network adapted feedback or implementable strategies for the defenders. We construct a class of parametric performance indices for which the defenders' strategies along with standard feedback Nash strategies of the attacker and target provides a network adapted feedback Nash equilibrium. We introduce a consistency criterion for selecting a subset (or refinement) of network adapted feedback Nash strategies, and provide an optimization based approach for computing them. Finally, we illustrate our results with numerical experiments.