DC bias optimization in intelligent DCO-OFDM Li-Fi systems using hybrid machine learning with hardware validation.

Area of Science:

• Wireless Communication
• Optical Networking
• Machine Learning Applications

Background:

• DC-biased optical orthogonal frequency division multiplexing (DCO-OFDM) is crucial for Light fidelity (Li-Fi) systems.
• Suboptimal DC bias in DCO-OFDM leads to performance issues like clipping noise and reduced optical power.
• Optimizing DC bias is essential for efficient Li-Fi system operation.

Purpose of the Study:

• To determine the optimal DC bias value for DCO-OFDM-based Li-Fi systems.
• To compare the effectiveness of different machine learning algorithms in predicting this optimal value.
• To validate the chosen ML model through hardware implementation.

Main Methods:

• Implementation and comparison of machine learning algorithms: hybrid linear regression with K-Nearest Neighbors (KNN) and hybrid polynomial regression with KNN.
• Evaluation of models using Root Mean Square Error (RMSE), Coefficient of Determination (R2), and Mean Absolute Percentage Error (MAPE).
• Hardware validation using an Arduino-based receiver for real-world signal data collection.

Main Results:

• Hybrid polynomial regression with KNN demonstrated superior performance in simulations, achieving RMSE of 0.18847, R2 of 96.908%, and MAPE of 9.271%.
• Hardware validation confirmed the model's effectiveness, yielding an RMSE of 0.296 and an R2 score of 81.37%.
• The results consistently showed the superiority of the hybrid polynomial regression with KNN model.

Conclusions:

• Hybrid polynomial regression with KNN is an effective technique for predicting the optimal DC bias in DCO-OFDM-based Li-Fi systems.
• The model's robustness was confirmed through both simulation and real-time hardware implementation.
• This approach offers a promising solution for enhancing Li-Fi system efficiency and performance.