eedavidwu/mimo-nos

Source Code for My Master Thesis in Umich; https://arxiv.org/pdf/2206.15065.pdf

MIMO-NOS

Source Code for My Master Thesis in Umich, "Learning-Based Near-Orthogonal Superposition Code for MIMO Short Message Transmission", submitted to possible journal publication.

Introduction

Apply deep learning based joint coding and modulation scheme termed as NOS code (Near-Orthogonal Superposition code) to the multi-input multi-output (MIMO) system for short packet transmission. Simulation results show the proposed scheme outperforms maximum likelihood (ML) MIMO detection combined with a polar code with CRC-assisted list decoding by 1 - 2 dB in various MIMO systems for short (32 - 64 bit) message transmission.

The flowchart of the MIMO-NOS:

Dependency

• Python (3.6.9 or above)
• numpy
• pytorch 1.10.0+cu102
• visdom

Reproduce

Basically, one can reproduce the PER results following two steps:

Learn a good codebook

Perform end-to-end training to obtain good codebook ${\cal C}$. E.g., to transmit a 64-bit packet in a 4x4 MIMO system, one should set the hyperparameters in 'mimo_simu.py' as $(V=8, M=256, Nt=Nr=4)$. Then type

$ python3.6 mimo_simu.py

to train the neural encoder and decoder. You are able to get a '.pth' file (neural network model), which will be used to construct the codebook for the second step.

Hand-crafted Looped-Kbest decoding

We design a hand-crafted looped-kbest decoder to decode the received signal. Sorting is used to determine the decoding order of the $V$ layers. Per-layer sorting is faster, while the Per-branch sorting achieve better PER performance.

To run the looped-kbest decoder, change the hyperparameters to be consistent with that in 'mimo_simu.py', type

$ python3.6 looped_kbest_bs.py

for Per-branch sorting, while run looped_kbest_ls.py for Per-layer sorting.

Baseline

Polar with maximum-likelihood (ML) MIMO detection serves as the baseline. We select Polar with CRC because of its superior performance in short block length. Moreover, the optimal ML detector ensures a good baseline.

Run the matlab code mypolar.m to obtain the PER curve.

Performance

PER performance for 64-bit packet transmission under 4x4 MIMO.

Moreover, the proposed MIMO-NOS is able to process with faster speed compared to the baseline.

Acknowledge

I've graduated successfully as a master student in the EECS department of the University of Michigan. Many thanks to my supervisor Hunseok and the group members, Mingyu, Chinwei and Changwoo!