Study On Data Fusion Algorithm Of Rocket Flight Multi-information Measurement

Multi-source information fusion is a method conducting information coming from multiple sources with multi-level, multifaceted, multi-level processing to generate new and meaningful information. Take measurement data of rocket flight path as the research object, multi-source information processing method is utilized to achieve complementary information, enhance system reliability, improve system survivability, and to get more accurate orbital data of rocket flight data.Basic principles of multi-source information fusion and its functional and structure models are presented in the thesis. After introducing its recent works, the current problems and future directions are analyzed. The processing content of the rocket orbit measurement data and the existed problems are thoroughly researched by combining the characteristics of the rocket orbit measurement data and measurement data processing method.As different sampling intervals, different observation coordinate system caused by the different measuring equipments, geographical location and other factors, the time and space alignment algorithms are studied to align the diverse and observational data to uniformed time and space fusion center. In the occasion of too large abnormal values in measurement data, the data matching check algorithm based on trajectory parameters is used to remove the outliers of long stretches.A state estimation algorithm of the rocket flight path under single station measurement is studied. The thesis mainly researches extended Kalman filter in nonlinear tracking systems. The extended Kalman filter is prone to scatter when the model of the system is inexact. To solve it, an improved extended Kalman filter algorithm with fading factor is proposed. The fading factor restricts the length of the data which processed by the Kalman filter. And we use"one step"approximation algorithm to calculate the suboptimal fading factor. The proposed filter algorithm has strong tracking ability, with a modest computational complexity and better suppression of the filtering divergence.In addition, the structure model in the data fusion and the data fusion algorithm of fusion center are studied in the thesis. Based on the structure model of general distributed data fusion, a dynamic distributed multi-level fusion model is proposed. This model reduces the uncertainty of system and simplify the fusion structure. Furthermore, fusion algorithm of the fusion center is deeply researched. Due to high computation of matrix-weighted fusion and low accuracy of scalar-weighted fusion, diagonal matrix-weighted fusion algorithm is adopted because it has proper computation and accuracy. This algorithm reduces the burden of online computation and improves the accuracy for processing real-timely.At last, the thesis summarizes the whole researched works, and points out the further research direction in this field.