This paper presents research on inverse optimal control in non-cooperative differential games. An inverse optimization framework for the identification of an unknown objective function of a game’s participant is introduced. The identified objective function yields the same trajectories as the ground truth objective function. We apply the presented methods in a human-machine cooperation scenario, where identification of human behavior is indispensable for automation design. The framework’s methods are evaluated through simulation of a driver cooperating with a driving assistance system based on a realistic vehicle dynamics model. The identification results are highly accurate and calculation performances motivate real-time applications.