Improving Smartphone Dynamic PPP by Adapting State and Measurement Covariance With TDCP-Aided Detection and Innovations t-Test

Smartphone-based location-based services (LBS) increasingly demand high horizontal positioning accuracy with reliability. However, the performance of smartphone precise point positioning (PPP) remains limited by two primary factors. First, due to the mass-market and compact design of smartphones, low-cost antennas and receivers introduce multipath effects and clock-reset errors, complicating the differentiation between reliable and unreliable measurements. Second, conventional constant-acceleration kinematic models cannot precisely capture rapid changes in motion, such as during acceleration, deceleration, or turning, leading to disturbances in the predicted state covariance in dynamic PPP. To address these issues, this paper explores the application of an adaptive robust quality control strategy for smartphone dynamic PPP, in which both state and measurement covariance are adaptively adjusted through dual detection mechanisms. Time-differenced carrier-phase (TDCP) observations, derived from robust least squares estimation, are employed to construct a learning factor for identifying model disturbances, enabling adaptive updating of the predicted state covariance matrix. To enable a statistically rigorous t-test on the standardized innovations for outlier detection, matrix projection is first employed to mitigate clock-reset effects and restore the normality of the innovations. The t-test then allows for the adaptive adjustment of the measurement noise covariance, improving filter robustness under dynamic conditions. Tests on datasets from typical expressway and urban complex scenarios demonstrate significant improvements in positioning accuracy. Horizontal root mean square error decreased by up to 27%, and 50th, 68th, and 95th percentile statistics improved by maximums of 32%, 33%, and 21%, respectively. These results indicate that adapting both state and measurement covariance through TDCP-aided detection and innovations t-tests effectively enhances the reliability and robustness of smartphone dynamic PPP in challenging dynamic scenarios.