| Abstrakt: |
Smartphones have become the mainstream terminals in the field of location services due to their low cost, portability, and ubiquity. In highly dynamic situations, the challenging urban environment causes the received Global Navigation Satellite System (GNSS) signals change rapidly and complexly, which brings great difficulties to error modeling. The estimation of GNSS measurement errors needs to consider multi-dimensional data, and the Kalman filter and graph optimization method cannot model the impact of the surrounding environment on GNSS position estimation. Therefore, we propose a method that utilizes machine learning algorithms to correct smartphone GNSS positioning errors. Specifically, the positioning error is predicted by the backpropagation neural network model optimized by the genetic algorithm that takes into account the carrier-to-noise density ratio, Doppler, position dilution of precision, and coordinate and velocity information obtained by pseudorange single-point positioning (SPP). Kinematic experimental results in four typical urban environments show that the proposed method can reduce the root mean square of the 3D positioning error from 25.08 m to 3.86 m. Compared with SPP, Kalman filter, and graph optimization methods, the positioning accuracy of this method is improved by 84.61%, 81.33%, and 74.59%, respectively. The proposed method has high potential for practical smartphone navigation and positioning applications. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink