/GNN-RNN-Based-Trajectory-Prediction-ITSC2021

Primary LanguagePythonMIT LicenseMIT

GNN-RNN-Based-Trajectory-Prediction-ITSC2021

This repo contains the code for the paper 'Graph and Recurrent Neural Network-based Vehicle Trajectory Prediction For Highway Driving' at ITSC-2021.

Data Pre-processing

Select the vehicles which have change their lanes only once through out the study area.

python once_lc_veh_selector.py

Preprocess data for the single trajectory prediction task.

python stp_data_pre.py

The dataset is managed by

stp_gr_dataset.py

Models

The proposed two channel model with GNN & RNN is implemented in stp_gr_model.py

The proposed interaction-only one channel model is implemented in stp_g_model.py

The proposed dynamics-only one channel model is implemented in stp_r_model.py

These models are based on a base model implemented in stp_base_model.py

Training

Train the two channel model:

python train.py

Citation

If you find this repo helpful in your work, please cite our paper as:

@inproceedings{mo2021graph, 
  title={Graph and Recurrent Neural Network-based Vehicle Trajectory Prediction For Highway Driving}, 
  author={Mo, Xiaoyu and Xing, Yang and Lv, Chen}, 
  booktitle={2021 IEEE International Intelligent Transportation Systems Conference (ITSC)}, 
  pages={1934--1939}, 
  year={2021}, 
  organization={IEEE} 
}

Thank you!