Atomic clock prediction algorithm: random pursuit strategy

Abstract

The present study proposes a novel prediction algorithm named 'random pursuit strategy'. It contains a predictor ensemble consisting of several predictors, each operating in a subspace of the original sample data space. The prediction is calculated by combining the outputs of the individual predictors using a weighted average. The frequency data of cesium clocks and hydrogen masers was predicted using the Kalman filter predictor and random pursuit strategy. The proposed algorithm demonstrates preferable capability in some cases. This could have beneficial applications for system controls in some areas.

A Pseudo Random Pursuit Strategy for Atomic Clocks

“…Although the Kalman filter approach is still the most efficient way for the clock real-time prediction and the theoretical time complexity of it is around O k 3 , RPS and PRPS have their advantages mentioned in section II. Such as, for minor jumps [29], RPS may have a greater potential advantage than the Kalman filter to reduce prediction errors due to anomalous clock behavior.…”

“…As a result, the RPS or PRPS is proposed to deal with the clock data prediction influenced by anomalous behaviors, especially for which perturb only a single (or a few sparsely distributed) reading(s) in the time series. Although the robust least square fitting or the median filter is a common method to deal with jumps, we emphasize that RPS can handle multiple forms of anomalous behaviors including minor jumps [29]. The robust least square fitting or median filter is not helpful for minor jumps.…”

“…In general, three parameters-phase, frequency, and frequency drift-can be used to describe the readings of Cs-clocks and H-masers [29], [30] about the modified Julian date (MJD). The prediction of the clock phase is the main emphasis of this example.…”

“…The simulation procedure with similar parameters [29] is used in this work. The frequency difference of an atomic clock relative to its reference is expressed as the formula:…”

“…As in the previous research, the random pursuit strategy (RPS) was used to reduce the influence of anomalous clock behaviors on its predictability [29], [30]. It was derived from Bagging [31], [32], which is a classical method of combinatorial optimization theory.…”