A Two-Stage Scheme for Target Localization in Multi-BS Cooperative ISAC Systems

Abstract

Multi-base station (multi-BS) cooperative sensing has emerged as a promising approach for achieving highaccuracy localization in sixth-generation (6G) integrated sensing and communication (ISAC) systems. In this paper, a two-stage localization scheme is proposed for multi-BS cooperative ISAC networks. First, a signal subspace-based global refinement orthogonal matching pursuit (SSGR-OMP) algorithm is developed for the high-accuracy joint extraction of multidimensional target information. By exploiting subspace information and continuousdomain gradient descent, the proposed method reduces parameter estimation errors in multi-target scenarios and mitigates off-grid bias. Based on the estimated parameters, a symbol-level data fusion localization method is further proposed by leveraging the phase consistency of intermediate symbols. Specifically, a joint weighted error function is constructed by incorporating symbol-level weights derived from the Cramér-Rao lower bound (CRLB) and data-level weights determined by the received signalto-noise ratio (SNR). The final target location is obtained via an iterative shrinkage search to minimize this error function within a constrained neighborhood. Simulation results demonstrate that the proposed scheme improves localization accuracy by approximately 57% and 60% compared with monostatic sensing and traditional data-level fusion, respectively, thereby validating its effectiveness for multi-BS cooperative localization.