Neighborhood Selection-Based Distributed Maximum Correlation Accumulation Direct Position Determination

Abstract

Existing distributed direct position determination algorithms with single-hop transmission face convergence issues under low signal-to-noise ratio (SNR) environment. In addition, these algorithms employ the conventional neighborhood all-selection (AS) strategy, where each sensor communicates and shares information with all neighboring sensors. In large-scale sensor networks, each sensor may have a substantial number of neighbors. Receiving and processing data from the entire neighborhood leads to significant communication overhead, high computational load, and slow convergence speed. To address these issues, this article proposes a distributed direct position determination algorithm with single-hop transmission. A cost function for distributed localization is derived from the classical centralized direct position determination, leveraging the correlation of received signals between sensors. To maximize the cost function, each sensor iteratively updates its estimate based on the diffusion strategy and the gradient ascent method. Two neighborhood selection strategies are proposed to select neighbors for each sensor. Data are only received and processed from the selected neighbors, resulting in reduced communication and computation within the sensor network. Experimental results demonstrate that the proposed algorithm maintains good convergence even under low SNR environment. Compared to the AS strategy, the proposed neighborhood selection strategies reduce communication overhead and computational burden of the sensor network, while enhancing the convergence speed of the algorithm.