pip install -r requriements.txt
All the files (NN dataset, NN, trajectory, etc.) generated during training and control
To generate the dataset for LSTM training
python data_generation_dynamic.py
To train one LSTM controller for each module
python main.py --mode train
To test
python main.py --mode test
To train one LSTM controller for the four-module robot
python main.py --mode train
To test
python main.py --mode test
To control the robot to generate [task] motion (including 'spiral', 'obs', and 'straight')
python ctrl.py --task [task]
To generate experimental results figures and gif,
python show.py --task [task]
python plot.py --task [task]
Share the same principle wit 2_ctrl_large_LSTM
We ask that any publications which use this repository cite as following:
@article{chen2024novel,
title={A Novel and Accurate BiLSTM Configuration Controller for Modular Soft Robots with Module Number Adaptability},
author={Chen, Zixi and Bernabei, Matteo and Mainardi, Vanessa and Ren, Xuyang and Ciuti, Gastone and Stefanini, Cesare},
journal={arXiv preprint arXiv:2401.10997},
year={2024}
}