IRIS employs an off-policy Deep Reinforcement Learning (DRL) framework, effectively circumventing the limitations observed in on-policy approaches and enhancing learning adaptability and efficiency.
Status: This project is part of ongoing research. The related paper is currently under submission.
To run the project, use the following command:
cd irlids_codebase
python project_dqn_new.pyThis project is based on the work presented in the following reference:
@InProceedings{cao2022catnipp,
title = {Context-Aware Attention-based Network for Informative Path Planning},
author = {Cao, Yuhong and Wang, Yizhuo and Vashisth, Apoorva and Fan, Haolin and Sartoretti, Guillaume},
booktitle = {6th Annual Conference on Robot Learning},
year = {2022}
}
@InProceedings{IPP_IRLIDS,
title = {IRIS: An Information Path Planning Method Based on Reinforcement Learning and Information-Directed Sampling},
author = {Ziyuan Liu and Yan Zhuang and Peng Wu and Yuanchang Liu},
year = {2024}
}