Machine Learning-Based 5G-and-Beyond Channel Estimation for MIMO-OFDM Communication Systems.

Area of Science:

• Wireless Communications
• Signal Processing
• Machine Learning

Background:

• Accurate channel estimation is vital for wireless network performance.
• Deep learning shows promise in improving communication reliability and reducing complexity in 5G+ networks.
• Least squares (LS) estimation is common but suffers from high error rates.

Purpose of the Study:

• To propose a novel deep learning-based channel estimation architecture.
• To improve upon the accuracy of traditional LS channel estimation methods.
• To evaluate the framework's effectiveness in 5G-and-beyond multiple-input multiple-output (MIMO) systems.

Main Methods:

• Developed a deep learning framework for channel estimation.
• Simulated a MIMO system with a multi-path channel profile and Doppler effects.
• Explored various artificial neural network architectures, including bidirectional long short-term memory (BiLSTM).

Main Results:

• The proposed deep learning framework significantly outperformed traditional channel estimation methods.
• BiLSTM demonstrated superior channel estimation quality and the lowest bit error ratio.
• The architecture is generalized for arbitrary numbers of antennas and neural network designs.

Conclusions:

• Deep learning-based channel estimation offers a superior alternative to conventional techniques.
• BiLSTM is a highly effective architecture for advanced wireless channel estimation.
• The proposed method enhances system performance in 5G-and-beyond mobile environments.