Conical Localization via Modified Polar Representation: A Unified Framework for Robust 3-D Positioning with 1-D Sensor Arrays

Abstract

This paper presents a unified framework for robust three-dimensional (3-D) source localization using a network of sensors equipped with one-dimensional (1-D) linear arrays. While such arrays offer practical advantages in terms of cost and size, existing localization methods suffer from a fundamental limitation: their performance degrades significantly as the source moves into the far-field, a common challenge known as the thresholding effect. To address this issue, we reformulate the localization problem in the modified polar representation (MPR) coordinate system, which parameterizes the source location using its azimuth, elevation, and inverse-range. We have developed a constrained weighted least squares (CWLS) estimator, which is subsequently transformed into a tight semidefinite programming (SDP) problem via semidefinite relaxation, enhanced with additional constraints to improve accuracy. Simulation results demonstrate that the proposed estimator attains the Cramer-Rao lower bound (CRLB) for both angle and inverse-range estimation in near-field scenarios. More importantly, it maintains this optimal performance in the far-field, substantially outperforming state-of-the-art methods, which exhibit significant error at large ranges. The proposed solution thus provides a reliable, unified localization system that is effective irrespective of the source range.