Online Inverse Identification of Noncooperative Dynamic Games

In this paper, the inverse optimization problem of discrete-time noncooperative dynamic games is investigated, where the cost functions of the dynamic games are required to be recovered based on the observation of open-loop Nash equilibrium trajectories. The solution to this problem allows the generation of a model to infer the interconnection pattern of a network system consisting of many dynamic agents. The inverse optimization problem (or the recovery of the cost functions) is solved by an online method which is derived from the necessary conditions of Nash equilibria. The proposed online calculation method can handle either finite or infinite-horizon problems, and the state and control input data are processed recursively. The performance of the proposed method is illustrated through numerical examples.