RACNN algorithm for robust adaptive beamforming
Fulai Liu,Liwen Feng,Hao Qin,Long Li,Yajie Gao,Ruiyan Du
DOI: https://doi.org/10.1007/s11276-024-03873-4
IF: 2.701
2024-12-04
Wireless Networks
Abstract:Via an atrous convolutional neural network (ACNN), a robust adaptive beamforming algorithm is presented for solving the output signal to interference plus noise ratio (SINR) degradation problem in the case of steering vector (SV) mismatch, named as RACNN algorithm. In the proposed algorithm, an ACNN model firstly is constructed to predict beamforming weight vectors, which consists of an atrous convolutional black, a max-pooling layer and a fully-connected block. Specially, three atrous convolution layers are introduced into the atrous convolutional black to improve the prediction accuracy of beamforming weight vector, which is achieved by enlarging the receptive field without increasing the parameters of the network. Then, a label is generated by estimating SV and reconstructing interference plus noise covariance matrix. By using the label, better robustness of the RACNN algorithm can be achieved. Finally, the robust adaptive beamforming weight vectors can be rapidly and precisely output by the well-trained ACNN model. Simulation results show that the proposed algorithm has better beamforming performance, for example: 1) it increases the robustness to the SV mismatch with DOA error ; 2) it achieves higher output SINR when SNR varies in [–10 dB,30 dB]; 3) it has better real-time performance and its speedup ratio reaches 98.95%.
computer science, information systems,telecommunications,engineering, electrical & electronic