In this repository, we release models from the papers
- [Learning-accelerated A* Search for Risk-aware Path Planning], which will be present in SCITECH 2024
This study aims to enhance the efficiency of solving the Constrained Shortest Path (CSP) problem. We introduce a novel approach that employs a transformer-based neural network, specifically a decoder architecture, to generate heuristic values for the A* algorithm. This methodological innovation results in a significant acceleration of the A* algorithm's performance, outpacing the traditional Manhattan heuristic in terms of speed.
train
python train.py
test predict, you need to change the args.model_path into the path you saved the model(end with .pt) first
python visualization.py
formal evaluation, you need to change the arg.model_path into the path you saved the model(end with .pt) first
python evaluate_model.py
@inproceedings{xiang2024learning,
title={Learning-accelerated A* Search for Risk-aware Path Planning},
author={Xiang, Jun and Xie, Junfei and Chen, Jun},
booktitle={AIAA SCITECH 2024 Forum},
pages={2895},
year={2024}
}