STOCHASTIC MODELLING OF PSEUDOLITE CLOCK ERRORS USING ENHANCED AR METHODS

摘要

In a pseudolites based real-time locating system, AR techniques are applied to model the stochastic behavior of the pseudolite clock's errors. Low-cost clocks are characterised by noise types like Flicker and Random Walk, which cannot always be modelled with rational transfer functions. Moreover, due to the white measurement noise some parts of the stochastic process can be obscured. A stable model to predict the clock deviations is obtained from the AR parameters of a process smoothed by a Kalman filter. The obtained model is used to predict the clock's behaviour for different time intervals. This model is integrated in the position calculation process in the pseudolite system. A performance comparison of the proposed scheme with an approach based on a Markov model is performed using measurements from a real-world measurement system. It is shown that the proposed scheme can predict the clock deviations with an accuracy of picoseconds leading to location errors of few centimeters.