Research On Quantum Stochastic Filter And The Computation Of Its Parameters

In the field of stochastic filtering, the state transition model of the observe signal is usually unknown. Designing an intelligent filter which is not based on any priori knowledge is very significant to reduce stochastic diffusion. Quantum filter can implement model-free stochastic filtering by using Schr?dinger equation as the universal state transition equation. Therefore, the study of quantum filtering algorithm and improving its applicability have important application value.Targeting to improve accuracy, stability and applicability of quantum filtering algorithm, this thesis mainly focuses on:(1) In order to improve the stability and the accuracy of the quantum filter, this paper proposes to preprocess the observe signal by using a variable variance Gaussian kernel function. The principle of improving the stability and the accuracy of the quantum filter is detailedly analyzed. Finally, the simulation comparison improved quantum filter with traditional RLS shows the advancement of improved quantum filter in accuracy, adaptivity and flexibility.(2) To calculate the parameters of the quantum filter quickly, parameters are divided into system-parameters which are independent from observed signal and non-system parameters which are only based on the observed signal. The function formulas relying on the frequency of input sinusoid signal and the input SNR are established for non-system parameters. Finally, the simulation comparison filter whose parameters are calculated with the new method discussed in this paper with filter whose parameters calculated with the traditional genetic algorithm shows effectiveness and efficiency of the new parameter calculation method discussed in this paper.(3) To ensure the accuracy of the filter in the non-sinusoidal input cases, the short-time Fourier transform is used to estimate the frequency and SNR of the observe signal. Then, the non-system parameters are can be estimated dynamically. Finally, the simulation comparison quantum filter whose non-system parameters are fixed with the quantum filter whose non-system parameters are updated dynamically shows the superiority and shortage of the non-system parameters on-line update method discussed in this paper.(4) To use the spatial correlation property of the vector observe signal, this paper proposes a no feedback quantum filter whose computational complexity is much lower. Then, the no feedback quantum filter is expanded into two dimensional quantum filter. Finally, the simulation comparison the two dimensional quantum filter with two independent single dimensional quantum filter shows merits and drawbacks of the two dimensional quantum filter.