Action-Conditioned Traffic Scene Prediction for Interactive Planning

Autonomous vehicles must be able to understand the movements of surrounding vehicles and predict how the future traffic conditions will be for planning a safe trajectory. During prediction, the action of autonomous vehicles should be considered, as it influences the movements of other vehicles sharing the same traffic scene and thus influences the future traffic flow. In this paper, we present a novel learning-based framework that forecasts a nearby traffic scene conditioned on the action of autonomous vehicle. Through experiments, we demonstrated that the proposed method can generate traffic scene which is more helpful to planning than that which do not consider the action of autonomous vehicle.