Dynamic Hypergames for Synthesis of Deceptive Strategies With Temporal Logic Objectives

In this paper, we study the use of deception for strategic planning in adversarial environments. We model interaction between agent (player 1) and adversary 2) as a two-player concurrent stochastic game which has incomplete information about agent’s task specification given temporal logic formula. During interaction, can infer intention from observations adapt its strategy so to prevent satisfying objective. To plan against such an adaptive opponent, must leverage knowledge adversary’s influence behavior thereby be deceptive. synthesize deceptive strategy, introduce class hypergame models that capture asymmetric, information. develop solution concept hypergames show subjectively rationalizable is maximizes probability logic. Such obtained by modeling opponent’s evolving perception objective integrating it into planning. This allows manipulate induce opponent taking actions benefit agent. demonstrate effectiveness our algorithm using robot motion examples with objectives design detection mechanism notify potential errors behavior. Note Practitioners—Many security defense applications employ mechanisms advantages. work presents game-theoretic framework strategies environments shows plays key role effective mechanisms. For cyber-physical security, practitioners specifying properties system analyze proposed methods.