Pedestrian Trajectory Prediction Based On Human-vehicle Interaction

Pedestrian trajectory prediction in traffic scene is an important part of intelligent vehicle and intelligent transportation system.At the same time,it is also an important basis for solving the problems of traffic congestion,intelligent vehicle driving decision and system control.In the traffic scene,the pedestrian’s trajectory will be affected by the surrounding pedestrians and vehicles,so it is important to study the social behavior of pedestrian trajectory to improve the accuracy of pedestrian trajectory prediction.The main work of this paper is as follows:Firstly,the DUT human-vehicle interaction data set was constructed,and the track sequence of pedestrians and vehicles in the traffic scene was obtained by using the correlation filtering method.Then,the Kalman filtering method was used to smooth the obtained data.Secondly,the fan-shaped human-human interaction neighborhood and the circular human-vehicle interaction neighborhood were designed to accurately capture the pedestrians and vehicles that interact with the predicted pedestrians,and the anti-collision function and direction attention function of human-vehicle interaction are defined as the weight of human-vehicle,human-vehicle social information,which further improves the accuracy of social information.Then,three different LSTM coding layers were used to encode the human-human,human vehicle interaction information and the current historical information of the predicted pedestrian.After that the human-human,human-vehicle interaction information was inputted into the attention module to get information that pedestrians pay more attention to.Finally,the filtered social information and the encoded pedestrian history track sequence are input into the LSTM prediction network to predict the pedestrian track.At last,the DUT human-vehicle interaction data set constructed by our group was used to verify the network proposed in this paper.The experimental results show that the method proposed in this paper can accurately predict the future movement trajectory of pedestrians in traffic and improve the accuracy of intelligent vehicle decision.