Kinematic GPS Batch Processing, a Source for Large Sparse Problems

Abstract

In kinematic observation processing the equivalence between the state space approach (Kalman filtering plus smoothing) and the least squares approach including dynamic has been shown (Albertella et al., 2006). We will specialize the proposed batch solution (least squares including dynamic), considering the case of discrete-time linear systems with constant biases, a case of practical interest in geodetic applications. A discrete-time linear system leads often to large sparse matrices, and we need efficient matrix operation routines and efficient data structure to store them. Finally, constant biases are estimated using domain decomposition methods. Simulated and real data examples of the technique will be given for kinematic GPS data processing.