Robust Long Baseline Underwater Target Localization Based on RANSAC

To address the degradation of traditional long-baseline (LBL) positioning accuracy caused by acoustic ranging errors in underwater environments, this paper proposes an improved LBL method based on Random Sample Consensus (RANSAC). Conventional LBL algorithms are susceptible to outliers in the presence of occasional ranging errors, leading to significant deviations in positioning results. The proposed RANSAC-LBL method iteratively filters inliers and constructs robust localization models to mitigate the impact of abnormal measurements. Simulation results demonstrate that when 1-2 sporadic errors exist in ranging data, the average positioning error of RANSAC-LBL is reduced by over 99% compared to traditional methods (from hundreds of meters to centimeters). Field experiments in Danjiangkou Reservoir further validate the effectiveness of RANSAC-LBL, achieving a mean absolute error (MAE) of 5.97 meters, which represents a 75% improvement over the traditional method's 23.13 meters. This approach provides an effective solution for high-precision underwater localization in complex environments.