Channel Prediction Based on BP Neural Network for Backscatter Communication Networks

Area of Science:

• Wireless Communication
• Sensor Networks
• Machine Learning

Background:

• Backscatter communication networks utilize ultra-low power sensors, generating large data volumes that necessitate increased network throughput.
• Existing rate adaptation methods struggle to simultaneously consider spatial and frequency diversity or incur high channel probing costs.

Purpose of the Study:

• To propose an efficient channel prediction scheme for backscatter networks to enhance network throughput.
• To address the limitations of existing methods by enabling simultaneous consideration of diversity and reducing channel probing expenses.

Main Methods:

• A two-part scheme involving a monitoring module and a prediction module.
• The monitoring module uses acceleration sensor data to track node movement and a link burstiness metric (β) to detect environmental changes, signaling the need for channel quality updates.
• The prediction module employs a back propagation (BP) neural network algorithm for forecasting future channel quality.

Main Results:

• High accuracy in predicting channel quality for backscatter networks.
• Significant improvement in network goodput was demonstrated through experimental implementation on readers.

Conclusions:

• The proposed channel prediction scheme effectively enhances backscatter network performance.
• The integration of sensor data and BP neural networks offers a promising approach for optimizing wireless sensor network throughput.