Learning the Pedestrian-Vehicle Interaction for Pedestrian Trajectory Prediction

TL;DR

The results show that the pedestrian-vehicle interaction influences pedestrian behavior, and the models using the proposed PVI extractor can capture the interaction between pedestrians and vehicles, and thereby outperform the compared methods.

Abstract

In this paper, we study the interaction between pedestrians and vehicles and propose a novel neural network structure called the Pedestrian-Vehicle Interaction (PVI) extractor for learning the pedestrian-vehicle interaction. We implement the proposed PVI extractor on both sequential approaches (long short-term memory (LSTM) models) and non-sequential approaches (convolutional models). We use the Waymo Open Dataset that contains real-world urban traffic scenes with both pedestrian and vehicle annotations. For the LSTM-based models, our proposed model is compared with Social-LSTM and Social-GAN, and using our proposed PVI extractor reduces the average displacement error (ADE) and the final displacement error (FDE) by 7.46% and 5.24%, respectively. For the convolutional-based models, our proposed model is compared with Social-STGCNN and Social-IWSTCNN, and using our proposed PVI extractor reduces the ADE and FDE by 2.10% and 1.27%, respectively. The results show that the pedestrian-vehicle interaction influences pedestrian behavior, and the models using the proposed PVI extractor can capture the interaction between pedestrians and vehicles, and thereby outperform the compared methods.

Figures (3)

• Figure 1: The overall framework of pedestrian trajectory prediction when considering the vehicles. The part inside the blue dotted polygon box is the proposed Pedestrian-Vehicle Interaction Extractor.
• Figure 2: The Pedestrian-Vehicle Interaction Module.
• Figure 3: The comparison of prediction results of LSTM, Social-LSTM and SI-PVI-LSTM. (a) The scenario where pedestrian A is turning right, avoiding the moving vehicles B, C, and D. (b) The scenario where pedestrian A has turned right and keeps walking straight, avoiding moving vehicle B and parked vehicle C. (c) The scenario where pedestrian A is crossing the road, interacting with vehicle B that is slowing down and waiting. The legends: obs denotes for observed paths of pedestrians; gt refers to the ground truth of predicted trajectories of pedestrians. veh_obs refers to the observed vehicle trajectories, and veh_future stands for the future trajectories of vehicles during the prediction time. lstm refers to the LSTM model, s-lstm refers to Social-LSTM; si-pvi-lstm denotes our proposed Social Interaction and Pedestrian-Vehicle Interaction LSTM model.