Kalman Filter-based Single-baseline GNSS Data Processing without Pivot Satellite Changing

• Main Authors: ZHANG Baocheng, YUAN Yunbin, JIANG Zhenwei
• Format: Article
• Language: zho
• Published: Surveying and Mapping Press 2015-09-01
• Series: Acta Geodaetica et Cartographica Sinica
• Subjects: GNSS single-baseline; Kalman-filter; between-receiver single-difference; double-difference ambiguity; pivot satellite; frequency transfer
• Online Access: http://html.rhhz.net/CHXB/html/2015-9-958.htm
• ISSN: 1001-1595; 1001-1595

Single-baseline global navigation satellite system (GNSS) data are able to be processed into a batch of parameters such as positions, timing information as well as atmospheric delays. The applications of relevance, therefore, consist of relative positioning, time and frequency transfer and so forth. To achieve real-time capability, these parameters are usually estimated by means of Kalman-filter. Moreover, the reliability of these parameters can be further strengthened by forming and then successfully fixing a set of independent double-differenced (DD) integer ambiguities. For this purpose, the filter function model is commonly set up based on the DD observation equations (DD filter model). In order to preserve the continuity of the filter, DD filter model needs to explicitly refer to another pivot satellite once the previous one becomes invisible. This thereby implies that, before being predicted to the next epoch, the former filtered DD ambiguity vector has to be “mapped” with respect to the newly-defined pivot satellite. In addition to that, the estimated receiver phase clocks using DD filter model may soak up distinct between-receiver single-differenced (SD) ambiguities belonging to different pivot satellites and would thereby be subject to apparent “integer jumps”. In this contribution, SD observation equations involving estimable DD ambiguity parameters are alternatively selected as the filter function model (SD filter model). Our analyses suggest that, both DD and SD filter models are equivalent in theory, but differ from each other as far as their implementations are concerned. Typically, for SD filter model, no effort should be made to map DD ambiguities, thus implying less intensive computational burden and better flexibility than DD filter model. At the same time, receiver phase clocks determined by SD filter model are free from “integer jumps” and thus are particularly beneficial for frequency transfer.