Deception in Learning of Games: A Closed-Loop Stackelberg Game Analysis

In this paper, we consider deception mechanisms in the learning of games. We consider a scenario in which certain Intelligent Players (IPs) attempt to modify the score function values of the benign agents to those desired by the IPs. In response, these agents modify their score values to prevent the effect of the IPs’ added input. This is analyzed as a Stackelberg game, under the closed-loop information structure (that is, all players can observe the whole state trajectory). This game is studied in a continuous time setting with tracking as one of the objectives in the cost function. We demonstrate the efficacy of our framework through simulations.