The accuracy and precision of clock synchronization plays an important role in synchronized pseudolite systems. Though the atomic clock is a favorable choice for time-keeping application, the crystal oscillator is obviously more desirable in view of the cost. Due to the lower stability and accuracy of crystal oscillators, additional approaches for frequency transfer and time offset calibration are required for a pseudolite network, and approaches based on wireless links are very desirable for their convenience. However, due to the immobility of pseudolites and complicated wireless environments, the measurements unavoidably contain constant errors, such as the multipath error and the tropospheric delay. From a global perspective, if pseudolites are able to receive signals from multiple other pseudolites, there will be more information available to mitigate the influences of errors mentioned above, which is a motivation of this work. The measurements are collected from the whole network, and the clock offsets of all pseudolites are estimated together. Moreover, in our approach, the pseudolite use all received signals to compute the frequency correction for itself. Also, the robustness is considered and relevant improvements are discussed in this paper. In harsh environment, the Least-Square estimation may be severely deteriorated by abnormal measurements of large bias. We put forward an alternative method by using robust estimator. By feat of the measurements from the whole network, Lorentzian estimator is involved to automatically mitigate their bad influences and improve the robustness of our approach The paper describes the main procedure of the network-based synchronization approach. Its performance is verified based on results of Monte-Carlo simulations. Simulation results show that when no abnormal signals are involved, the Lorentzian estimator will obtain a sub-optimal estimation with a slightly worse performance than Least-squares (LS). However, it significantly outperforms LS estimation under the influence of abnormal signals. We also discuss the selection of parameters for the Lorentzian estimator and the influences on the final performance. The precision of the proposed frequency transfer approach which combines signals from multiple pseudolites is verified as well.
展开▼
