CSI Compression and Prediction
AI for CSI feedback compression and CSI prediction enhancements
These examples show AI techniques for channel state information (CSI) feedback compression and CSI prediction enhancements in 5G wireless communications systems. Use them to step through a workflow that includes data generation, data preparation, deep neural training, compression, system testing, and deployment.
Topics
Introduction
- AI-Based CSI Feedback (5G Toolbox)
End-to-end workflow for examples exploring channel state information (CSI) feedback compression techniques using artificial intelligence (AI) in 5G wireless communication systems. (Since R2026a)
Data Generation
- Generate MIMO OFDM Channel Realizations for AI-Based Systems (5G Toolbox)
Generate channel estimates to train an autoencoder for CSI feedback compression and temporal channel prediction. (Since R2026a)
Data Preparation
- Preprocess Data for AI-Based CSI Feedback Compression (5G Toolbox)
Preprocess channel estimates and prepare a data set to train an autoencoder for CSI feedback compression. (Since R2025a) - Preprocess Data for AI Eigenvector-Based CSI Feedback Compression (5G Toolbox)
Preprocess channel estimates and prepare a data set to train an autoencoder for eigenvector based CSI feedback compression. (Since R2026a) - Preprocess Data for AI-Based CSI Prediction (5G Toolbox)
Preprocess channel estimates and prepare a data set to train a gated recurrent unit (GRU) channel prediction network. (Since R2026a)
Model Training
- Train Autoencoders for CSI Feedback Compression (5G Toolbox)
Compress CSI feedback using an autoencoder neural network in a 5G NR communications system. (Since R2022b) - Train Transformer Autoencoder for Eigenvector-based CSI Feedback Compression (5G Toolbox)
Train an autoencoder neural network with a transformer backbone to compress downlink CSI over a clustered delay line (CDL) channel. (Since R2026a) - CSI Feedback with Transformer Autoencoder (5G Toolbox)
Design and train a convolutional transformer deep neural network for CSI feedback by using a downlink clustered delay line (CDL) channel model. (Since R2024b) - Optimize CSI Feedback Autoencoder Training Using MATLAB Parallel Server and Experiment Manager (5G Toolbox)
Accelerate determination of the optimal training hyperparameters for a CSI autoencoder by using a Cloud Center cluster and Experiment Manager. (Since R2024a) - Offline Training and Testing of PyTorch Model for CSI Feedback Compression (5G Toolbox)
Train an autoencoder-based PyTorch® neural network offline and test for CSI compression. (Since R2025a) - Online Training and Testing of PyTorch Model for CSI Feedback Compression (5G Toolbox)
Train an autoencoder-based PyTorch neural network online and test for CSI compression. (Since R2025a) - Train PyTorch Channel Prediction Models (5G Toolbox)
Train a PyTorch neural network for channel prediction by using data generated in MATLAB®. (Since R2025a) - Train PyTorch Channel Prediction Models with Online Training (5G Toolbox)
Enable real‐time adaptation to time‐varying wireless channels by generating each training batch in MATLAB on-the-fly to train a PyTorch GRU channel prediction network online. (Since R2026a)
Model Testing
- Test AI-based CSI Compression Techniques for Enhanced PDSCH Throughput (5G Toolbox)
Measure physical downlink shared channel (PDSCH) throughput in a 5G New Radio (NR) system, with a primary focus on AI-based compression methods for CSI feedback. (Since R2026a) - CSI Feedback Compression for 802.11be Using AI (WLAN Toolbox)
Use a k-means based AI/ML technique to compress CSI feedback in an 802.11be SU-MIMO beamforming scenario. (Since R2025a)
Deployment
- CSI Feedback with Autoencoders Implemented on an FPGA (Deep Learning HDL Toolbox)
This example demonstrates how to use an autoencoder neural network to compress downlink channel state information (CSI) over a clustered delay line (CDL) channel. (Since R2024b)
