Abstract:

In terms of the characteristics of instantaneous velocity pulses, piecewise constant empirical accelerations and periodic perturbation accelerationmodels, effective computational methods are proposed for calculating the sensitivity matrices related to model parameters respectively. Meanwhile,design matrices and normal equation matrices are expressed in the form of different partitioned sub-matrices. The computational amount of numerical integration and data storage of effective methods grows with O(n), butconventional methodsdo with O(n 2). The GRACE-A on board GPS observation data has been used to evaluate the computionalefficiency of normal equationby use of the proposed methods.Results show that for instantaneous velocity pulses and piecewise constant empirical accelerations, the computational efficiency of modeling normal matricesis increasedonce when ionosphere-free linear combination of pseudorange observations is used, and the efficiency is increased 3.5 times when ionosphere-free linear combination of pseudorange and carrier phase observations is used. However, there is no obvious improvement for periodic perturbations because the number of parameters is small compared with the other two models.

Key words: Reduced dynamic model；Instantaneous velocity pulse；Piecewiseconstant empirical acceleration；Periodic perturbation, Partitioned sub-matrix