| As one of the greatest science and technology achievements in twentieth Century, global navigation satellite system (GNSS) has been widely applied ranging from geodesy and geodynamics to atmospheric detection. In the coming years, more and more new GNSS satellites will be available, for example, the Chinese Beidou (BDS) and European GALILEO systems are developing, the American GPS system is in its modernization. Meanwhile all the new satellites provide at least three frequencies services. All these bring great challenges to the current GNSS data processing strategies.From one side, almost all the current GNSS products are derived from the ionosphere-free combination or difference methods, however, these methods are not capable to fulfill all the requirements of the multi-constellation and multi-frequency environment. First of all, in order to implement the complementary of different GNSS systems and to provide consistent corrections for various applications, data at all frequency bands and from all systems should be processed together at the observation level, where inter-system and inter-frequency biases must be handed properly. And then, to make full use of the raw information, raw measurements instead of combined or differenced observations must be utilized. In this paper, a unified data processing model based on the raw measurements for multi-constellation and multi-frequency is studied in detail, data from any GNSS system and at any frequency bands could be processed following this method.From another side, the problem of low computational efficiency has not been solved well in GNSS data processing. Even if only for the GPS data, it is hard or impossible to solve networks with hundreds of stations using current combined or differenced data processing methods. What’s worse is that the raw data processing method is more time-consuming than the combined or differenced methods, which makes it rather hard to be used for daily routinely processing even with only about 100 multi-GNSS stations. Although subnetwork strategy could relieve part of the computation burden, however, it could not provide a rigorous solution. Therefore, this paper is aimed to propose a new efficient strategy which could be used not only to improve the data processing efficiency for large-scale GNSS network, but also to accelerate the raw data processing. The detail contributions of this dissertation include:(1) The current data processing methods and its challenges are introduced firstly; then, the raw data processing model is studied in detail, including the methods to separate the highly correlated parameters, such as inter-system/inter-frequency biases, DCB, clocks and ionosphere delay parameters. The ambiguity-fixing methods are also studied when no constraints are applied to ionospheric delay parameters.(2) The consistency between frequencies are investigated and discussed for GPS and BDS. Experiments show that the residuals of GPS is abnormally larger than that of L1 and L2, and this kind of residuals behave similarly for the same station-satellite pair in a few days, which are probably caused by some unknown system biases. However, no similar residuals are found at any BDS frequency bands. Additionally, the DCB of L5 also show a good stability with monthly STD less than 3 cm.(3) The products of raw data processing are accessed by comparing them to those derived from ionosphere-free combination processing method. The results confirm a very good agreement of orbits, clocks, troposphere delays and inter-system biases in the case of using dual-frequency data.(4) A new efficient GNSS data processing strategy is proposed, where PPP ambiguity-fixing methods and Carrier-range concept are combined. This new strategy includes six steps:a) precise orbits and clocks are determined using a global network; b) UPDs are estimated using the float ambiguities in last step; c) PPP and ambiguity-fixing is carried out for each station; d) phase observations are converted to carrier-ranges at stations base; e) repeat c) and d) until the processing of all stations are finished. f) the integrated processing is carried out with carrier-ranges. In the final step, no ambiguities or only a few unfixed ambiguities need to be estimated, thud the efficiency could be improved greatly.(5) The stability of the UPDs are analyzed. Experimental data show that the UPD of a Block IIA satellite often has a jump of about 0.3 to 0.4 cycles after an eclipse, while the UPDs of other satellites are rather stable. And after removing the jumps, the UPDs of those eclipsed BLOCK IIA satellites are also very stable. Therefore we recommend that an additional UPD parameter should be estimated after an eclipse for Block IIA satellites.(6) Carrier-range is found to be capable to improve the data continuity. In the case of using carrier-ranges, all data segments belonging to a satellite-station pair are connected continuously, if their UD-ambiguities are fixed correctly. That means in the carrier-range method all stations "track" the satellites continuously even a satellite is not visible.(7) The efficiency problem for large-scale GNSS network data processing is solved after applying the new strategy. The experiments show that, the new strategy takes only 14 minutes to solve a network with about 460 stations, while the traditional strategy takes about 82 minutes. And the required computation time of the new strategy increase almost linearly with the increase of the number of stations, while that of traditional method increase exponentially with the increase of the station number.(8) Efficient high-rate clock estimation procedures for post and real-time cases are designed. Experiments show that, for the post processing, it takes only 32 min to calculate the 30s clock for 24 hours data with a network of about 250 stations, while more than 200 min are needed for the traditional method; for real-time processing, each epoch takes less than Is when 200 stations are adopted, whereas traditional method needs about 10 s.(9) The efficiency of raw data processing is also improved greatly after adopting the new strategies. Experiments show that, the new strategy shortens the processing time from 97 min to 31 min for a GNSS network with about 100 stations when no ionospheric constraints are imposed. It also shown that, even after imposing the ionospheric temporal and spatial constraints, the new strategy could also reduce the processing time from 387 min to 206 min. |
PDF Full Text Request