In this paper, we study the VPPP (Very Precise Point Positions) algorithm based on GR (GNSS Regression) models by using multiple antennas for estimating baseline vectors. The multiple antennas are disposed with solid geometrical distances which provide the constraints of the estimated parameters, but both antennas' positions are unknown. By using the baseline-vector length constraints, we show the VPPP algorithms based on the double-difference (DD) observables by each pair of two antennas, and apply the algorithm for estimating baseline vectors and GNSS Gyro. Finally we show the experimental results of the estimation algorithms for the baseline vectors and Euler angles.