General Framework for Multi-Sensor Multi-Target Localization: Sparse Solution and Iterative Data Association

The multi-sensor multi-target localization (MS-MTL) faces significant challenges, including data association ambiguities and computational complexity of processing the measurements from multiple sensors. In this article, we propose a general framework for MS-MTL that can be utilized in any arbitrary localization system. The framework consists of three stages: sparse representation and solution, data association, and multiple single-target localization. In the first stage, a sparse representation approach provides rough estimates of target locations, which does not achieve the Cramér-Rao lower bound (CRLB), most probably. In the second stage, the data association is performed by leveraging the rough locations to resolve ambiguities among targets based on maximum likelihood (ML) criteria. Finally, the third stage employs a quasi-Newton algorithm to refine individual target locations, using the initial estimates from the first stage as starting points for optimization. Furthermore, the association and localization stages can be repeated multiple times using the final locations as the next rough locations to refine results. We have evaluated the effectiveness and computational efficiency of the proposed framework using a multiple-input multiple-output (MIMO) system that relies on bistatic range measurements.