This repository contains the training and inference code used in our paper Integrating kinematics and environment context into deep inverse reinforcement learning for predicting off-road vehicle trajectories. This work proposes a inverse reinforcement learning based framework that infers the reward structure and forecasts the vehicle's motion.

click to view

We recommend using conda to install dependencies with the environment.yml provided in this repository.

conda env create -f environment.yml
source activate vehicle_motion_forecasting

You can also use pip to install dependencies with the requirements.txt provided.

pip install -r requirements.txt

We provide the trained weights and example data for inference. Please check demo.ipynb.

jupyter notebook demo.ipynb

Training examples will be made available later after we open source the dataset.

Please consider citing the corresponding publication.

@inproceedings{zhang2018integrating,
  title={Integrating kinematics and environment context into deep inverse reinforcement learning for predicting off-road vehicle trajectories},
  author={Zhang, Yanfu and Wang, Wenshan and Bonatti, Rogerio and Maturana, Daniel and Scherer, Sebastian},
  booktitle={Conference on Robot Learning},
  pages={894--905},
  year={2018}
}