COMPARISON OF LEAST SQUARES AND KALMAN FILTER SOLUTIONS FROM DIFFERENT IVS ANALYSIS CENTERS

Creative Commons License

TANIR KAYIKÇI E., KARAASLAN Ö.

23th EVGA Working Group Meeting, Gothenburg, Sweden, 14 - 19 May 2017

  • Publication Type: Conference Paper / Summary Text
  • Volume:
  • City: Gothenburg
  • Country: Sweden
  • Karadeniz Technical University Affiliated: Yes

Abstract

 

IVS Analysis Centers apply various statistical methods, namely Least-Squares (LSQ) method, the Kalman Filter (KF) method, the Square-Root Information Filter (SRIF) and the Least-Squares Collocation (LSQC) method and consider the behaviour of stochastic parameters in different approaches.  The majority of geodetic VLBI analysis softwares uses LSQ, e.g., CALC/SOLVE software, Vienna VLBI Software (VieVS) and OCCAM. QUASAR and OCCAM use LSQC.

Kalman filters or square-root information filters are applied in OCCAM and SteelBreeze. Eventhought Kalman Filter is not one of the commonly used techniques in VLBI analysis software, it has some advantages to determine short-term random variations in the estimation of tropospheric delays and clocks which might affect accuracy of estimated parameters accuracies. In Least-Squares (LSQ), parameters are described as constant through different measurement epochs. LSQ estimation supposes that the parameters that we want to estimate are constant for all observation equations in the problem. Nevertheless we can have the case that certain parameters in the same problem might have variations based on the time, atmosphere or any other causes. However, in Kalman Filter estimation procedure, parameters can have variations at each epoch and their behaviours can be desribed statistically so this procedure allows the estimation of instantaneous changes. Additonaly, with the Least-Squares estimation method; each observation requires the computation of a multidimensional matrix inverse. Computations with the Kalman Filter method are simpler and faster, so the method is very convenient when a number of parameter changes must be quickly analyzed.

In this study, we first consider comparison of KF and LSQ solutions from different IVS analysis centers to some ideas about the procedures which can be implemented in Kalman Filter output to make it combinable with LSQ results are given for VLBI intra-technique combination.