Deep learning aided multi-source passive 3D AOA wireless positioning using a moving receiver: A low complexity approach

Abstract

Passive localization of multiple sources while the receiver is in motion is a fundamental problem in wireless communication with practical applications in various fields such as acoustic event localization and object tracking. In this study, we propose a novel method for three dimensional (3D) angle of arrival (AOA) localization using a moving receiver with a limited number of sensors. In the proposed method, we record signals from sources in a series of time windows. Then we estimate the elevation and azimuth AOAs of sources along with their pairing by using the proposed three low-complexity deep learning (DL)-based AOA estimators. Following the initial AOA estimation, a low-complexity off-grid based refinement technique is employed to further enhance the accuracy of the estimated elevation and azimuth AOAs based on the maximum likelihood (ML) criterion. Next, a multi-source 3D-localization algorithm is proposed to estimate the positions of sources across the recorded time windows. The simulation results validate the effectiveness and efficiency of the proposed method. The results highlight that the DL-based AOA and location estimators outperform recent studies, while also demonstrating robustness in the face of practical imperfections, such as situations where the receiver’s position or direction is uncertain during movement. Our complexity analysis verifies that the proposed method outperforms existing methods in terms of speed and the number of computations. This enhanced computational efficiency makes the proposed method highly attractive and practically significant for application in various fields.