Sub-metre Positioning with Smartphone Global Navigation Satellite System Measurements in User Environments

The ubiquity of smartphones catalyzes various smartphone-based Internet of Things (IoT) applications, among which smartphone positioning that uses Global Navigation Satellite System (GNSS) observations to provide user position plays an important role. The noisy smartphone GNSS measurements from signal obstructed environments and embedded antennas prevent users from obtaining reliable and accurate positioning solutions. With a correct understanding of the measurement characteristics and advanced positioning techniques such as Real-Time Kinematic (RTK), Precise Point Positioning (PPP), and PPP-RTK, this dissertation aims to achieve sub-metre- or decimetre-level positioning accuracy with smartphone GNSS measurements in realistic environments. In this dissertation, a novel smartphone range error derivation method is proposed, as well as an improved cycle slip (CS) detection method. New stochastic modelling and measurement outlier detection methods are developed. And smartphone ambiguity resolution (AR) is conducted with a newly proposed candidate selecting strategy.

Smartphone measurements exhibit significantly higher noise levels as compared to those from geodetic hardware, in which the range errors are more correlated with the carrier-to-noise density ratio (C/N0). Moreover, range errors are environment-dependent, and they behave differently when, e.g., the smartphone is mounted on the vehicle roof versus the dashboard. Considering that the conventional Doppler CS detection method is sometimes not applicable due to inconsistent time-differenced carrier phase (TDCP) and Doppler clocks, a novel modified Doppler method is proposed and validated with simulated cycle slips. With proposed Doppler cycle slip detection method, all simulated cycle slip combinations are detected. In terms of quality control methods, the improved stochastic model and novel outlier detection method outperform the conventional C/N0-based weighting and a fixed residual rejection threshold, showing a percentage improvement from 38% to 54% for PPP horizontal positioning errors within 1 metre. A novel smartphone RTK wide-lane (WL) partial AR strategy is proposed, for which the static results show an improvement of 83% in horizontal when WL AR is conducted compared to the underlying float solution, and the positioning accuracies can reach 6.8, 2.9, and 11.5 cm in E, N, and U, respectively. The proposed algorithm is validated with kinematic datasets collected in real-life environments, in which the time series of horizontal positioning errors exhibit less variation than the float solutions, represented by smaller positioning errors. Moreover, with fixing WL ambiguities, the horizontal positioning performance has improvements ranging from several centimetres to up to 8 decimetres depending on the related environments. the biggest improvement of 0.79 m is noticed for 95th percentile horizontal positioning errors under suburban environments. Improvements on the above-mentioned aspects show potential on achieving decimetre-level positioning performance with smartphone embedded antennas in realistic environments.