Incomplete Observations of the Dynamic Location Algorithm
|School||Central South University|
|Course||Geodesy and Survey Engineering|
|Keywords||missing data fitting combination method dynamic positioning gray interpolation sub-optimal kalman filter method|
In the GPS dynamic positioning, those loss of datas are called missing data which caused by variety of causes between the observation data. Because it not need to store large amount of observation datas in the process of solving, and when the new observation datas obtained directly calculate the value of filtering parameter to achieve the purposes of real-time observation data processing, so Kalman filtering theory has become the most important method of data processing in dynamic positioning. Dynamic positioning data obtained time series data, these datas processing method are very strict on data integrity, and we could not get the high-precision dynamic positioning solution duo to the long time losing of data. So some of observation data loss will seriously affect the data processing results and reduce the dynamic positioning accuracy, and these cases are usually unable to make up for quality control.At present, the method for handling missing data are mainly EM algorithm and multiple imputation and multi-sensor fusion methods. EM algorithm deal with missing data for dynamic positioning observation, it can not estimate the missing data and can only estimate the model parameters. Imputation is filling the missing observations, and a variety of methods have emergenced currently. But there have a lot of problems among the process of deal with missing data on dynamics observation because of the most filling methods are not proposed for the dynamic positioning observation data and not took the characteristics of dynamic positioning real-time processing into account. Multi-sensor fusion method has not yet formed a unified theory, and the robust nature of the problem are not well resolved.This paper conducts the following research based on the above:1. Given the research background of algorithm in dynamic localization when the observed data is missing, and analysised the current research situation of missing data in geodesy data.2. Compared with several commonly used methods for missing data problems, analysised the basic principles among them, and proposed the lack of these methods applied to deal with dynamic positioning missing data.3. For the deficiency of current used in commonly data processing methods, taking the dynamic Kalman filter algorithm in a wide range of positioning applications into account, and combine with the characteristics of real-time processing on dynamic positioning, then proposed three method to treatment dynamic positioning missing observation which based on kalman filter.4. Validated the algorithm by simulation example and real GPS data, compared the adjustment with the standard Kalman filter result, and then given the application of the algorithm and weaknesses.