MIMO detection using an ALR-CNN in SDM transmission systems.

Area of Science:

• Optical communication systems
• Signal processing
• Machine learning in communications

Background:

• Multiple-input multiple-output (MIMO) detectors are crucial for modern communication systems.
• Mode division multiplexing (MDM) offers increased data capacity but requires sophisticated detection.
• Existing MIMO detectors face performance limitations in complex optical transmission systems.

Purpose of the Study:

• To design and implement a novel MIMO detector by integrating adaptive learning rate (ALR) with convolutional neural network (CNN).
• To evaluate the performance of the proposed ALR-CNN MIMO detector in an MDM optical transmission system.
• To compare the effectiveness of the ALR-CNN detector against conventional detection algorithms.

Main Methods:

• Development of a new MIMO detector architecture combining ALR and CNN.
• Implementation of the ALR-CNN detector within an MDM optical transmission testbed.
• Performance evaluation through training and testing signal accuracy and Bit Error Rate (BER).
• Comparative analysis against traditional MIMO detection techniques.

Main Results:

• The proposed ALR-CNN MIMO detector achieved 100% training and test accuracy.
• The ALR-CNN detector demonstrated superior performance compared to conventional MIMO detectors.
• The system successfully reached ideal Quadrature Phase Shift Keying (QPSK) BER levels.
• A minimum Signal-to-Noise Ratio (SNR) difference of approximately 9.5 dB was observed at a BER of 10^-3.

Conclusions:

• The ALR-CNN approach provides a highly accurate and effective solution for MIMO detection in MDM optical systems.
• This deep learning-based detector significantly enhances communication system performance.
• The proposed method offers a promising advancement over traditional MIMO detection algorithms.