Recurrent Neural Network Based Link Quality Prediction for Fluctuating Low Power Wireless Links

Area of Science:

• Wireless communication networks
• Signal processing
• Machine learning applications

Background:

• Link quality prediction is crucial for wireless networks.
• Existing methods often fail to capture temporal dynamics and link fluctuations.
• This leads to inaccuracies, particularly in variable link conditions.

Purpose of the Study:

• To propose a more effective link quality prediction method, RNN-LQI.
• To address the limitations of existing methods in handling temporal correlations and link fluctuations.
• To improve the accuracy of link quality prediction in low-power wireless links.

Main Methods:

• Utilizing Recurrent Neural Network (RNN) to predict Link Quality Indicator (LQI) series.
• Leveraging the short-term memory of RNN to mine inner relationships within LQI series.
• Employing a fitting model to evaluate link quality based on LQI and Packet Reception Ratio (PRR).

Main Results:

• RNN-LQI demonstrates superior performance compared to existing methods across various link qualities.
• Prediction error is reduced by at least 14.51% for moderate fluctuations and 13.37% for sudden changes.
• The method effectively handles link fluctuations due to LQI's higher resolution in transitional regions.

Conclusions:

• RNN-LQI offers a more accurate and robust approach to link quality prediction.
• The method is particularly suitable for low-power wireless links characterized by frequent fluctuations.
• RNN-LQI enhances the reliability of wireless communication systems in dynamic environments.