5GChannelPrediction

Implementation of the LSTM-based approach from Deep Learning based Wireless Channel Prediction: 5G Scenario (Varshney et al., 2023). This project predicts 5G mmWave channel responses in real-time, bypassing the need for pilot signals.

---

🎯 Project Overview

This research focuses on building and validating a lightweight LSTM-based model that calculates future wireless channel coefficients (H) and critical mobility parameters by learning from historical time-series measurements.

• Pilot-free Prediction: Exploits temporal autocorrelation to reduce network overhead and increase spectral efficiency.
• Mobility Awareness: Extends channel prediction to real-time mobility metrics (TX-RX distance and delay spread) for network-level optimization.
• Reproducible Research: Follows the architecture-exact parameters of the original paper with verified quantitative outcomes.

---

📊 Dataset & Configuration

The project uses measurement-based statistical channel models simulated with NYUSIM at 28 GHz.

Parameter	Value	Feature	Description
Carrier Frequency	28 GHz	H_real	Real part of complex channel coefficient
Scenario	Urban Macro (UMa)	H_imag	Imaginary part of channel coefficient
Environment	LOS	Distance	TX-RX separation (100–200m)
Bandwidth	800 MHz	Delay Spread	RMS multipath delay spread (0.8–1.8μs)

---

🔁 Implementation Pipeline

The execution sequence is structured into three distinct phases:

Phase 1: Simulation & DNN Baseline (main.ipynb)

Generates 28 GHz channel data and trains a baseline DNN (500-250-120 units) to estimate coefficients from transmitted/received pairs. This confirms the practical limitations of Dense networks in noisy 5G scenarios.

Phase 2: LSTM Forecasting (lstm.ipynb)

Trains a 3-layer LSTM (512-256-128 units) on historical coefficients. This model predicts future H values with high accuracy, enabling predictive beamforming and resource allocation.

Phase 3: Mobility Parameter Extension (additional.ipynb)

Applies the LSTM methodology to forecast dynamic mobility metrics (Distance and Delay Spread), demonstrating the model's generalized applicability for 5G resource optimization.

The mobility parameter prediction (Phase 3) is a proof-of-concept for testing methodology. The values used are simulated and intended for demonstration purposes; they do not accurately reflect real-world 5G mobility dynamics.

---

🚀 Getting Started

1. Prerequisites

Ensure you have a Python 3.10+ environment ready.

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Run Sequence

jupyter notebook notebooks/main.ipynb        # Run Simulation & Baseline
jupyter notebook notebooks/lstm.ipynb        # Train Channel Prediction Model
jupyter notebook notebooks/additional.ipynb  # Perform Mobility Analysis

---

📑 References & Documentation

🔬 Research & Reports

• Original Paper: Research PDF - Varshney et al.
• Technical Report: Full Project Report - Comprehensive architecture and result analysis.

📚 Detailed Walkthroughs

• Stage 1: Data Preparation & DNN Baseline
• Stage 2: LSTM Architecture & Training
• Stage 3: Mobility Parameter Analysis