To improve pseudolite positioning accuracy in the indoor multipath-contaminated environment, a multipath error correction model based on the user space position is formulated, which utilizes the spatio-temporal variation characteristics of indoor pseudolite carrier multipath. To be specific, the double-differenced carrier multipath errors of the static sampling points are extracted, and the three-dimensional coordinates of these points are taken as the input and the multipath error as the output. The error model is trained by the support vector regression based on the radial basis kernel function, and the optimal hyperparameters of the model are achieved by the leave-one-out cross-validation method. On the basis of this model, the double-differenced carrier observation equations are modified continuously through iteration. The position solutions can approximate the real coordinates to the greatest extent. Thus, the multipath is mitigated. The static relative positioning experiments in the strong-multipath-contaminated indoor environment show that the overall horizontal positioning accuracy after multipath correction is improved to the centimeter-decimeter level. The vertical accuracy within 1 m. This method can be implemented without system transformation, which is suitable for pseudolite high-precision positioning in the structured indoor environments.