V2X Latency Prediction
V2X Latency Prediction V2X Latency Prediction V2X Latency Prediction V2X Latency Prediction V2X Latency Prediction V2X Latency Prediction V2X Latency Prediction

Machine Learning

Author

Tech stack

Python Python PyTorch PyTorch Wandb Wandb Shapley Shapley

V2X Latency Prediction

📜 Description

A deep learning system for predicting network latency in Vehicle-to-Everything (V2X) communication scenarios over 4G/5G networks. The project uses LSTM models to forecast latency 8 seconds ahead, incorporates Explainable AI (XAI) techniques for model interpretability, and includes a design of a Zero-Touch Service Management (ZSM) pipeline.

⚙️ Features

  • Real production dataset – Uses a real-world production dataset with additional feature engineering
  • 8-Second Horizon Forecast – Forecast V2X network latency up to 8 seconds ahead
  • Involves Regression and Classification – Predicts continuous latency values and classifies network quality states
  • 4G/5G Support – Works with both 4G and 5G network scenarios
  • Federated Learning – Distributed training across multiple vehicles while preserving data privacy
  • Compares common LSTM architectures – Evaluates Seq2Seq, Encoder-Decoder, and standard LSTM models for performance
  • Explainable AI (XAI) – Model interpretability and feature importance analysis (SHAP)
  • ZSM Pipeline – Design for autonomous network operation