On Feasibility of Multiantenna Spectrum Sensing under User Mobility: An Experimental Learning Approach

In cognitive radio (CR) networks, the detection of available bands in high-mobile wireless systems still remains a challenging task due to channel dynamics such as Doppler effects and multipath propagation. This study experimentally evaluates the feasibility of a learning-based spectrum sensing (SS) approach using a multiantenna software-defined radio (SDR) platform. The proposed SDR testbed recreates the user mobility by emulating time-varying wireless environments. Our experimental measurements show the feasibility to adopt a learning approach to overcome the detection performance of modelbased detector under different conditions of the user mobility, such as the Doppler effect or the number of channel taps. Concretely, experimental findings indicate that an artificial neural network (ANN) significantly improves the detection performance relative to a generalized likelihood ratio test (GLRT), under a low signal-to-noise ratio (SNR) regime. These results validate the effectiveness of learning-based detection in mobile wireless scenarios and suggest its applicability in spectrum access systems requiring robust real-time adaptation.