Human trajectory prediction and generation using LSTM models and GANs

作者：

Highlights：

• New deep neural network models are proposed for trajectory prediction.

• LSTM and GAN1 models are used for unimodal predictions, GAN3 model for multimodal.

• Metrics are proposed for normalizing errors for more consistent comparisons.

• New dataset are proposed with low linearity and a high diversity.

摘要

•New deep neural network models are proposed for trajectory prediction.•LSTM and GAN1 models are used for unimodal predictions, GAN3 model for multimodal.•Metrics are proposed for normalizing errors for more consistent comparisons.•New dataset are proposed with low linearity and a high diversity.

论文关键词：Trajectory generation,Trajectory prediction,LSTM,GANs

论文评审过程：Received 16 July 2020, Revised 8 April 2021, Accepted 27 June 2021, Available online 3 July 2021, Version of Record 10 July 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108136