Python 3.8.0
torch == 2.0+cu118
cuda == 11.8
cudnn == 8.8.1
python mptest.py example.csv
When executing the script, a CSV file containing various test vectors must be provided.
Each row in the file represents a set of design parameters, primarily including the number of channels, code length, and target PSL.
Upon execution, a folder named with WD followed by the timestamp will be created, and the currently running code will be copied into this folder for reproducibility.
Additionally, a CSV file named with DL followed by the timestamp will be generated, documenting the execution time and optimization results for each set of parameters.
The results for each parameter set will be saved in a folder prefixed with Job.
The pyWaveform.mat file will contain the optimized waveform, while trainloss.mat will record the metrics during the optimization process.
For Weighted PSL optimization, both G and E need to be specified. This means that the lags between G and E will be assigned the weight W during the optimization process.
For instance, if G is set to 10, E to 20, and W to 0.999, this indicates that sidelobes between lags 10 and 20 will be suppressed.
If you set G to 0 and E to N-1, the optimization will apply to all lags for PSL.
Take a look at the file lse_time_perf.m.
Please cite this paper in your publications if it helps your research:
@ARTICLE{10739961,
author={Cao, Jiahui and Sun, Jinping and Wang, Guohua and Zhang, Yuxi and Wang, Wenguang and Wang, Jun},
journal={IEEE Transactions on Aerospace and Electronic Systems},
title={Neural Network-Inspired Phase-Coded Waveform Design for MIMO Radar Based on Gradient Descent},
year={2024},
doi={10.1109/TAES.2024.3488687}}