LSTM-Based Prediction Method of Surrounding Vehicle Trajectory

Abstract

With the rapid development of autonomous driving, how to understand the behavior of targets around autonomous driving has become an important part of the autonomous driving system. As the most common location data, trajectory data is a key direction of research in intelligent transportation systems. The trajectory prediction task aims to predict the future trajectory of the target (such as pedestrians, vehicles and other traffic participants) based on the current or historical trajectory and environmental information. The trajectory prediction result is one of the important information for the automatic driving system to make advance decisions. This paper proposes an LSTM-based algorithm for predicting the trajectory of surrounding vehicles. This paper studies the trajectory prediction problem with Seq2Seq information as the core structure. First, the related types and concepts of the trajectory data set are introduced, and then the superiority of LSTM compared with traditional machine learning prediction algorithms is analyzed. Secondly, the core structure of the algorithm and the superior performance in interactive modeling are introduced. Finally, the advantages and disadvantages of the method are summerized, and an overview of the development trend of trajectory prediction is made.