Outlier Rejection for Wireless Positioning Accuracy

The patent extracts and removes outlier measurements (NLOS signals) from the set of all wireless link measurements. By identifying and eliminating these harmful measurements through statistical analysis and outlier rejection algorithms, the system retains only the reliable LOS measurements for positioning calculation, thus resolving the contradiction between quantity and accuracy.

The patent changes the parameter of measurement reliability by introducing confidence metrics and quality thresholds. Measurements are re-evaluated based on their consistency with expected geometric relationships and signal characteristics, transforming the raw measurement data into filtered, high-confidence positioning inputs that maintain accuracy while utilizing multiple wireless links.

The patent implements feedback mechanisms where positioning results are continuously validated against geometric constraints and measurement consistency checks. Outliers are identified through iterative refinement processes that compare expected versus actual measurements, allowing the system to self-correct and maintain high accuracy without requiring overly complex external intervention systems.

The patent applies partial outlier rejection by filtering only the most significant outlier measurements that clearly deviate from expected patterns, rather than attempting to perfectly classify every measurement. This selective approach achieves sufficient accuracy improvement while avoiding the exponential complexity growth that would result from exhaustive analysis of all possible measurement combinations.