Abstract
Multiple-input multiple-output technology is now widely used in mobile devices to improve communication throughput. While additional antenna elements are cheap, RF elements are expensive and lead to high power consumption. Hence, interest is magnified in so-called antenna selection, where only a few out of all antenna elements are chosen. However, most existing antenna selection methods are too complex to be implemented on mobile devices. Industries are encouraged to use the maximum norm method, where the antennas that provide the largest power are selected. In this study, we propose AsNet, an antenna selection technique based on the data-driven deep learning approach. Given large volumes of training data, AsNet can automatically learn to select the best antennas efficiently. Surprisingly, AsNet can achieve similar performance to brute force method and the complexity is only slightly higher than that of random selection.