Clustering Combined Weighted TDoA Localization for Outlier Suppression

Traditional TDoA-based localization methods cannot autonomously detect or discard large anomalies, leading to significant errors with unbalanced measurements. This study introduces a TDoA-based clustering combined weighted (CCW) method that leverages measurement consistency to identify anomalies in redundant measurements, using only those with normal errors to enhance localization accuracy in unbalanced scenarios. In 2-D localization, at least three base stations are needed to determine the position. With three stations, the CCW method simplifies the solution by transforming the coordinate system and resolving the co-linear problem. When more than three stations are present, the method generates multiple three-base-station combinations (TBSCs), solving each to obtain minimal measurement solutions. These solutions are clustered to find subsets that meet the clustering criteria. If all minimal measurement solutions diverge or fail to converge, the CCW method performs a systematic, exhaustive search to identify and exclude unreliable base stations until a valid aggregated class is found or only three stations remain. The final position estimate is the weighted average of the elements in this class, with weights based on the TBSC’s Cramer-Rao Lower Bound (CRLB). Simulation results show that under unbalanced noise conditions, especially with six- and eight-base stations, a threshold effect occurs. When the ratio of the abnormal error level to normal error level exceeds a certain level, the CCW method outperforms other techniques by detecting and removing abnormal measurements, thus improving localization accuracy by focusing on primarily accurate measurements. In real acoustic indoor localization experiments, the CCW method significantly outperformed other methods. In static tests, it achieved sub-0.36 m accuracy for 90% of estimated positions, and in dynamic tests, it closely matched real trajectories with an average anomaly rate of only 1.1%.