High-precision regional ionospheric interpolation is essential for achieving reliable positioning in network RTK. However, in challenging scenarios such as long baselines and highly active ionospheric conditions at low latitudes, ionospheric delays are difficult to accurately resolve through regional interpolation. As a result, the residual ionospheric interpolation error exhibits considerable uncertainty, which severely degrades the service performance of network RTK. Existing studies have highlighted the necessity of jointly optimizing both the server-side and user-side algorithms. In response, this paper proposes an enhanced network RTK approach—the ionospheric stochastic constraint method (ISC)—which integrates a server-side ionospheric residual interpolation uncertainty indicator with a user-side ionospheric weighting model. Experiments were conducted using data from a long-baseline reference station network in a low-latitude region, with an average baseline length of 104 km. Using the virtual reference station (VRS) method as a benchmark, the performance of the ISC-based network RTK algorithm was thoroughly evaluated. The results demonstrate that, compared to the VRS method, the mean value of the 90th percentile horizontal error under the ISC method was reduced from 5.6 cm to 1.9 cm, and the accuracy was improved by 66.1%. The mean value of the 90th percentile of vertical error decreased from 10.2 cm to 4.4 cm, with an improvement of 56.7%. The average daily fixed rate of ambiguity increased from 80.4% to 97.6%. In addition, the ISC method achieves a good balance between positioning accuracy and convergence speed, and can basically complete the initialization within 10 epochs, which can meet the timeliness requirements of real-time positioning.
To address the issue of long ambiguity resolution (AR) time for existing medium-to-long baseline reference stations, this paper proposes a rapid ambiguity fixing method with empirical ionospheric constraints. The method directly treats the L1 and wide-lane (WL) ambiguities as unknown parameters to be estimated and performs synchronous filtering in the ambiguity domain. Furthermore, empirical ionospheric constraint factors are derived from real measurement data, thereby providing reasonable and effective variance constraints for the weighted estimation of ionospheric parameters. While accelerating ambiguity convergence, the proposed method also simplifies the solution procedure. Experimental results show that, compared with the classical ionosphere-free (IF) model, the proposed method significantly improves the accuracy of the ambiguity float solution and the baseline initialization speed. The RMS of the ambiguity float solution bias is reduced from 1.15 cycles to 0.45 cycles, representing an accuracy improvement of approximately 60.8%. The average baseline initialization time is shortened from 51.2 epochs to 21.7 epochs, corresponding to a speed enhancement of about 57.6%. In addition, the proposed method also markedly improves the ambiguity fixing speed for newly risen satellites at low elevation angles: the average time required for ambiguity fixing is reduced from 50.1 epochs to 14.8 epochs, and the average required elevation angle drops from 23.4° to 15.7°.