when GPS is used to determinate the inter-satellite relative position and the precise dynamic model of inter-satellite relative movement cannot be obtained

the least square method is usually used in the post data processing

but in practice

the stochastic model used by LS method do not take into account the characteristic of heteroscedastic

space- and time-correlated of real GPS measurements.In order to solve this problem and improve the positioning precision

this paper proposes a new relative positioning method

which first founds the spline function model based on the continuity of the relative position

then eliminate the time correlation of GPS double-difference measurements based on epoch-difference theory

then uses MINQUE to solve variance-covariance components of de-correlated GPS measurements iteratively

finally uses LAMBDA to fix the integer ambiguities and determinates the relative position.The experiment simulation gets two conclusions:(1)spline function model can not only benefit eliminating time correlation of GPS measurements in dynamic condition

but also improve the relative positioning precision greatly because of its functions of saving the number of unknown parameters and smoothing noise;(2)the iterative stochastic modeling using MINQUE method can improve relative positioning precision more based on spline function model.