Skip to content

GOKULPRASANTH-M/1Bit-MIMO-DL-based-Precoding

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

1-Bit Massive MIMO Precoding Using Deep Learning

Overview

This project presents a deep learning–based 1-bit precoding framework for downlink massive MIMO systems. The proposed method learns an efficient mapping from Channel State Information (CSI) to 1-bit quantized precoding vectors, enabling low-power and hardware-efficient transmission suitable for next-generation 5G and 6G wireless systems.

Key Contributions

  • Deep learning–based 1-bit precoding for massive MIMO
  • Unsupervised training using a constructive interference–based loss function
  • Lightweight residual architecture with depth-wise separable convolutions
  • Robust performance under imperfect CSI and varying SNR conditions
  • Suitable for real-time and energy-efficient wireless communication systems

Model Architecture

The architecture consists of complex-aware preprocessing followed by efficient convolutional and attention-based residual blocks. A power-aware quantization layer ensures compatibility with 1-bit DAC constraints.

Main components:

  • Real–imaginary CSI preprocessing
  • Depth-wise separable convolution layers
  • Lightweight channel attention mechanism
  • Optimized residual blocks
  • 1-bit power-normalized quantization module

Dataset

The dataset simulates a realistic massive MIMO-OFDM downlink system.

  • Base station antennas: 32
  • Users: 4
  • Subcarriers: 64
  • Total samples: 120,000

Hardware impairments modeled:

  • 1-bit DAC quantization
  • Phase noise
  • IQ imbalance
  • Power amplifier nonlinearity

Training Setup

  • Optimizer: AdamW
  • Learning rate: 3e-4
  • Batch size: 16
  • Epochs: 40
  • Loss function: Constructive interference–based unsupervised loss
  • Training strategy: Chunked data loading for memory efficiency

Evaluation Metrics

  • Bit Error Rate (BER)
  • Signal-to-Noise Ratio (SNR)
  • Cosine Similarity
  • Mean Squared Error (MSE)

Test Performance

Metric Value
SNR ~18.2 dB
BER ~5.4%
Cosine Similarity ~0.87

Technology Stack

  • Programming Language: Python
  • Deep Learning Framework: PyTorch
  • Scientific Libraries: NumPy, SciPy
  • Tools: Git, Linux

Applications

  • Massive MIMO wireless systems
  • Low-resolution DAC transmitters
  • Energy-efficient 5G and 6G base stations
  • Hardware-constrained communication systems

Author

Gokulprasanth M
Bachelor of Engineering in Electronics and Communication Engineering

References

This project is based on recent research in one-bit precoding, constructive interference, and deep learning–based signal processing for massive MIMO systems.

About

Deep Learning Model Based 1Bit Multiple Input Multiple Output Precoding in Communication Networks like 4G,5G,etc..

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors