Radar-aided Beam Prediction based on SqueezeNet Network

Abstract

In this paper, a new radar-assisted beam prediction method is proposed based on SqueezeNet model for millimeter wave system. We use a large real data set to test our proposed radar aided optimal beam prediction model, DeepSense6G, which includes millimeter-wave beam training and radar measurement. In the test results section of this paper, we can see that our model has achieved a good balance between prediction accuracy and model complexity. In addition, the prediction accuracy of the first five beams of the proposed model can reach about 94%, and at the same time, it saves a lot of beam training overhead.