In the field of autonomous driving, anticipation of the dynamic environment is of great importance for the ego vehicle to make decisions and plan future paths in order to ensure safety and efficiency. This thesis presents a general goal point model for making predictions of vehicle motions around a moving ego vehicle. One or multiple goal points are selected based on a road graph and other environmental information. Vehicle predictions are then initialized from a probabilistic tracker and propagated via a motion model toward the goal point. This anticipation model is validated on a real-time dataset and evaluated against an open-loop, purely kinematic baseline model, demonstrating its predictive performance over a 1.5-second window in various scenarios.