Research Of Pointing Error Modeling Of The Satellite Laser Ranging Based On BP Neural Network

With the rapid development of optical, mechanical, electrical technology, satellite laser rangingtechnology has become more mature. On the aspect of data accuracy, both fixed station and mobilestation in the world enter subcentimeter era. In recent years, with high repetition rate laser technologywidely used, the domestic SLR stations all improve data quantity and accuracy. However, thedomestic SLR stations all have being done research in the fields of daytime ranging, blind tracking,unmanned observation, lunar laser ranging, and then improving pointing accuracy will be inevitable inthe near future.In the satellite laser ranging observation, improving the telescope pointing accuracy has theimportant effect of strengthening the ability of capturing and tracking targets. Especially, it is one ofthe most important indicators to obtain observation data, in the blind tracking and daytime ranging.Pointing error models which are used widely include spherical harmonic model, basic parametermodel and mount model. In detail, the model coefficients of the spherical harmonic model have nophysical meaning and the model is not stable. On the contrary, the basic parameter model is verystable, and its model coefficients have definite physical meaning. Compared with the basic parametermodel, mount model has higher pointing precision because it is an extension of former, but former ismore stable than latter.A modeling method based on the BP neural network, which is employed to pointing error of theSatellite Laser Ranging Telescope, is proposed. BP Neural Network is the one of most popular modelnowadays and multi-layer forward feed network by error back propagation algorithm. BP NeuralNetwork algorithm is applied to the pointing error model of mobile-satellite laser ranging not onlycan expand the field of application of neural network algorithm, but also has the great significance ofimproving the tracking accuracy of the satellite laser ranging system.This paper carries out the experiment by using three groups of observation data from Wuhanmobile Satellite Laser Ranging station and establishes the BP Neural network model, mount model,spherical harmonic model and the basic parameter model. It compares the four kinds of models bycalculating their root mean square error. The result shows that BP Neural Network can improve the pointing accuracy of the Satellite Laser Ranging Telescope, and is superior to the other three models.In addition, in order to analyze the distribution characteristic of pointing error, the article dividesevery group of observation sample from Shanghai station, Changchun station and Wuhan mobilestation two parts including training set and testing set based on the genetic algorithm, and models byusing the mount model. And this paper calculates the residual of elevation and azimuth angles of bypredicting the testing set sample. The experiment results show that residual are not necessarilyconform to the normal distribution and the genetic algorithm can optimize the model and improve thepointing error accuracy.Finally, taking into account the practicality of pointing error model and further improving themodel, this paper introduces the interface process of every model based the Matlab developmentenvironment.