The cut-in scenario is one of the challenging traffic situations for autonomous vehicles since it requires the ego car to not only avoid the static obstacles on the road but also to react quickly to the sudden behavior changes of the surrounding traffic participants. Traditional approaches commonly adopt a predict-then-plan architecture which decouples the Prediction and Planning modules thus lacking of interactions between them. A planning algorithm considering dynamic games theory is proposed in this paper to get multiple decisions when encountered with a cut-in scenario. Profit functions considering the state deviations and the lane change cost of the side car are designed.