Stepwise Goal-Driven Networks for Trajectory Prediction

Abstract

We propose to predict the future trajectories of observed agents (e.g., pedestrians or vehicles) by estimating and using their goals at multiple time scales. We argue that the goal of a moving agent may change over time, and modeling goals continuously provides more accurate and detailed information for future trajectory estimation. In this paper, we present a novel recurrent network for trajectory prediction, called Stepwise Goal-Driven Network (SGNet). Unlike prior work that models only a single, long-term goal, SGNet estimates and uses goals at multiple temporal scales. In particular, the framework incorporates an encoder module that captures historical information, a stepwise goal estimator that predicts successive goals into the future, and a decoder module that predicts future trajectory. We evaluate our model on three first-person traffic datasets (HEV-I, JAAD, and PIE) as well as on two bird's eye view datasets (ETH and UCY), and show that our model outperforms the state-of-the-art methods in terms of both average and final displacement errors on all datasets. Code has been made available at: https://github.com/ ChuhuaW/SGNet.pytorch.

Pedestrian Behavior Prediction Using Deep Learning Methods for Urban Scenarios: A Review

“…They extracted the features through embedding layers, and fed the features into deep learning structures for prediction. In addition to only trajectories, studies [25], [26], [44] encoded intermediate destinations from the trajectories and predict future trajectories conditioned on the destinations. For intention prediction, Zhao et al [66] used trajectories extracted from roadside LiDAR sensors to predict the crossing intention.…”

“…Recently, conditional variational autoencoders (CVAEs) with sequential encoders and decoders have been adopted to predict multi-modal distributions. The BiTraP [25], SGNet [26], CGNS [47] and DESIRE [55] used GRU encoder-decoders based on CVAE method for trajectory prediction with multi-modal goal estimation. Social-NCE [38] applied LSTM model based on the noise-contrastive estimation (NCE) methods [105] by introducing a social contrastive loss, namely the InfoNCE loss [106].…”