Channel Prediction Based on BP Neural Network for Backscatter Communication Networks

Abstract

Backscatter communication networks are receiving a lot of attention thanks to the application of ultra-low power sensors. Because of the large amount of sensor data, increasing network throughput becomes a key issue, so rate adaption based on channel quality is a novel direction. Most existing methods share common drawbacks; that is, spatial and frequency diversity cannot be considered at the same time or channel probe is expensive. In this paper, we propose a channel prediction scheme for backscatter networks. The scheme consists of two parts: the monitoring module, which uses the data of the acceleration sensor to monitor the movement of the node itself, and uses the link burstiness metric β to monitor the burstiness caused by the environmental change, thereby determining that new data of channel quality are needed. The prediction module predicts the channel quality at the next moment using a prediction algorithm based on BP (back propagation) neural network. We implemented the scheme on readers. The experimental results show that the accuracy of channel prediction is high and the network goodput is improved.