2D-CNN-Based AoA-ToA Estimation in Presence of Angle-Dependent Phase Errors Using Pico-cells

Many advanced location-based services and network optimization applications in the fifth generation (5G) mobile networks rely on accurate location awareness. In view of this, we investigate joint estimation of the angle-of-arrival and time-of-arrival of mobile user equipments using 5G pico-cell base stations. In practical scenarios, the measured phase of the channel state information (CSI) is distorted by angle-dependent errors induced by hardware impairments. To solve this problem, we propose a estimation algorithm based on two-dimensional (2D) convolutional neural network (CNN) in this work. Specifically, the spatial and subcarrier dimensions of the CSI are organized into the input matrix. Then, the features are extracted through the 2D convolutional layers. The 2D-CNN is further incorporated with a MUltiple-SIgnal-Classification (MUSIC)-based region decision method. As confirmed by numerical results, the proposed frame-work can effectively fit the non-linear angle-dependent array error and improve the accuracy of positioning.