Abstract Pedestrian trajectory prediction aims to use observed human historical trajectories and surrounding environment information to predict the future location information of target pedestrians. This research has important application value and can reduce the collision risk of autonomous vehicles under social interaction. However, traditional model-driven pedestrian trajectory prediction methods are difficult to predict pedestrian trajectories in complex and highly dynamic scenarios. In contrast, data-driven pedestrian trajectory prediction methods rely on large-scale data set platforms to better capture and model more complex pedestrian interactions, thereby achieving more accurate pedestrian trajectory prediction effects and becoming a popular choice for autonomous driving and robot navigation. and video surveillance and other research hotspots. In order to macroscopically grasp the research status and key issues of pedestrian trajectory prediction methods, the classification of pedestrian trajectory prediction technologies and methods is used as the starting point. First, the research progress of existing methods of pedestrian trajectory prediction is detailed and the key issues and challenges currently existing are summarized; Secondly, based on the modeling differences of pedestrian trajectory prediction models, existing methods are divided into model-driven and data-driven pedestrian trajectory prediction methods. At the same time, the advantages, disadvantages and applicable scenarios of different methods are summarized. Then, the methods used in pedestrian trajectory prediction tasks are summarized. The mainstream data sets are summarized and the performance indicators of different algorithms are compared; finally, the future development direction of pedestrian trajectory prediction is prospected.
Keywords pedestrian trajectory prediction; data-driven; social interaction; autonomous driving
0 Preface
Traffic safety issues have always been the focus of today's society, and a safe road environment is a necessary condition for autonomous vehicles to drive under social interaction. According to the "Global Road Safety Status Report" released by the World Health Organization (WHO), the number of global road deaths every year has reached an unprecedented 1.35 million. Among them, more than half of the accident victims are vulnerable road users (VRU), such as road pedestrians, cyclists and motorcyclists. Therefore, the road driving safety issues of the VRU group need to be focused on. In addition, pedestrians, as important participants in traffic scenes, are the biggest victims of traffic accidents. Reasonable reasoning and prediction of their future movement trajectories are important for autonomous driving decision-making and road planning.