The atomic clock prediction algorithm is a critical part of the atomic time scale system to ensure its stability and accuracy. Random pursuit strategy (RPS) has been verified on the prediction capability of hydrogen maser and cesium clock in our previous works. This method is applied to deal with data impacted by different noise types including hetero-variance white noises and jumps. Nevertheless, it is difficult to apply RPS to real-time clock prediction, owing to its computational complexity. To alleviate it, we further improved our original algorithm by simplifying the random grouping using a pseudo-random strategy. In our work, theoretical analysis and necessary simulations of the pseudo-random pursuit strategy (PRPS) are presented, and the experimental results show PRPS’s remarkable advantage in terms of operational efficiency. Compared to the original algorithm, it shows comparable accuracy and stability for prediction. PRPS takes only 1/p fitting time consumption as long as RPS starts from the second prediction. PRPS is faster, more efficient, and easier to employ when utilizing a clock predictor as the output of a system than RPS.
The Pseudo Random Pursuit Strategy (PRPS) has been proven to be an effective predictive algorithm for atomic clocks. In this paper, the Fast Pseudo Random Pursuit Strategy (FPRPS) is proposed to improve the computational efficiency of PRPS and reduce its time complexity. The core of FPRPS is to replace the inefficient repetitive calculation process in PRPS with an incremental updating process. The FPRPS is employed to predict the frequency difference of atomic clocks relative to UTC(NIM). The experimental results show that FPRPS can significantly improve processing speed without reducing predictive accuracy and stability, compared with PRPS. In experiments, the speed advantage is more than five times. The reduction in computing resource overhead indicates that FPRPS has a wider prospect of applications in engineering practice than PRPS.
The Recursive Pseudo Random Pursuit Strategy (RPRPS) is proposed to predict the frequency difference of atomic clocks relative to reference. It further improves the computational efficiency and reduces the time complexity. The core of RPRPS is to replace the refitting of the updated predictor and recalculating of the sum of square residuals of the unupdated predictors in a recursive process. In experiments, it is employed to predict the readings of the cesium clock and hydrogen maser relative to UTC (NIM). Compared with PRPS, the experimental results show that RPRPS reduces running time by about 70% without reducing predictive accuracy.
SummaryThis paper proposed a new weighting method for clock ensemble: redundant-subspace-weighting (RSW). In this approach, multiple overlapping subspaces are constructed in clock ensemble. By superimposition averaging of different subspace covariances, RSW could reduce the correlations between clocks located in the same laboratory. Preliminary simulation results show its potential capability to enhance the frequency stability of clock ensemble time scale.
Atomic clock frequency jumps directly influence the accuracy and reliability of timekeeping systems. The necessary corrections are typically implemented by postprocessing mutual comparison data between multiple atomic clocks based on the overly strict assumption that these atomic clocks are independent of each other. This paper describes the concept of a mirror clock, which enables atomic clock frequency jumps to be identified in real time without any assumptions. By comparing whether the real measured data and a corresponding mirror clock prediction fall within a confidence interval determined by the uncertainty of past physical clock data, atomic clock frequency jumps can be effectively identified and corrected. The results of several experiments using three hydrogen masers verify that the precision and recall of simultaneous jump identification reach 96.41% and 73.49%, respectively.