Abstract：We propose a GNSS instantaneous clock offset model construction algorithm considering parameter sparsity. The algorithm is constrained by adding L1 norm term(Lasso) to the least squares polynomial fitting, and the clock offset model is fitted by sliding modeling, that is, sliding Lasso fitting method. Using GFZ 300 s clock offset products as training data, the regularization parameters are obtained through 10-fold cross-validation, and the clock offset model is established. The satellite clock offset fitting accuracy and PPP accuracy are compared with the commonly used clock offset densification algorithm. The experimental results show that the sliding Lasso fitting method can effectively eliminate the influence of excessive jump of clock offset in local clock offset data during the fitting process, prevent the model from overfitting, and improve the fitting accuracy and PPP accuracy of the clock offset model.
