An Overview of the Motion Prediction of Traffic Participants for Host Vehicle

Abstract

The prediction dataset of trajectory, intention and maneuver is helpful to improve the performance and comfort of autonomous vehicle, and even it is the basis of safety. This paper focuses on the analysis, comparison and prospect of the prediction methods of surrounding participants in the traffic environment. Firstly, according to the decision processes, the paper organizes motion modeling and prediction approaches in three levels: physical trajectory, maneuver intention and interactive situation prediction. Further, different research methods are discussed from the aspects of prediction modeling: fundamental steps, road network model, models of participant vehicles, considered uncertainty, prediction results of driving behavior, etc. We also discuss the limitations of each kind of the prediction approaches. Finally, we point out some challenges about motion prediction of traffic participants.