Session E2: LEO for Positioning, Navigation, and Timing

Mixing Real and Simulated Observables to Assess the Performance of Hybrid GNSS/LEO-PNT Precise Positioning
Raul Orus Perez, Miguel Cordero Limon, Pietro Giordano, and Roberto Prieto-Cerdeira, European Space Agency ESA/ESTEC
Date/Time: Wednesday, Sep. 18, 4:46 p.m.

A number of initiatives aiming to deploy LEO-PNT systems are emerging all around the globe like Xona's Pulsar, Centispace and ESA's LEO-PNT. With the advent of different LEO-PNT constellation systems, a priori evaluation of their navigation performances is critical in order to properly tackle the system design. In particular, one of the benefits that LEO-PNT can bring is a faster convergence of precise positioning solution without the need of precise ionospheric corrections. In order to properly assess this potential advantage, simulation tools capable of assessing the performance of precise positioning algorithms exploiting both GNSS and LEO-PNT measurements (and the conditions under which such performances can be achieved) are needed.
Several simulators capable of performing end-to-end simulations to generate and process both GNSS and LEO-PNT observables have been presented. These tools are very versatile since usually the navigation engines work with full observations in RINEX format. However, there is the risk that the applied simulated models and the navigation engine models may differ and may add some error in the analysis that may not be due to the studied system. Moreover, one could argue that the error models used to generate simulated MEO (and LEO) observables could be not representative of reality with the risk of yielding optimistic or pessimistic results.
In this work, a methodology and tool supporting both fully-simulated and hybrid scenarios is presented. In the hybrid-scenario configuration, the combination of real observables from GNSS satellites in MEO collected during real test campaigns in the field and simulated LEO-PNT observables are processed. This allows to perform tests representative of real-life scenarios, and even reprocess GNSS data from previous test campaigns with the addition of the LEO-PNT component, in order to compare current GNSS-only performance and future GNSS/LEO-PNT expected performance. The tool also can generate basic urban scenarios as an azimuth-elevation mask that depend only in one parameter (ratio of width and height of the street, with typical ranges from 2 to 0.125, that can define most of urban situations), it can be complemented with the pedestrian user location with an additional parameter that ranges form 1 to 0.
The key for the generation of the hybrid data is a very simple idea. This is to reuse the position (critical in mobile platforms), receiver clock (and intersystem bias if need it) and tropospheric estimates computed from the real data. This can be done in postprocess applying any software tool and postprocess processing, such as forward/backward filters or/and integration with Inertial Measurement Unit (IMU). Then, the data for the new constellation is simulated at the level of the prefit residuals. Thus, only the error components are added to the clock and tropospheric estimates. These errors are, broadcast LEO clock error, LEO position error, Ionospheric error and, observation and receiver noise. Additionally, one could add cycle slips in the observations. For the Full simulation, one could reuse the same methodology by generating a synthetic receiver clock and tropospheric components and generate the same type of errors for the MEO constellation. Notice that all presented results are done with 1Hz data using uncombined PPP ambiguity float solutions.
Initial results with this methodology show how different error components for a large LEO constellation (for 304 satellites) on the model affect the convergence time on the user receiver. They show that the geometry of the simulated urban canyons (as an azimuth-elevation mapping on the observations) can affect the design of the LEO-PNT constellation (trying full constellations of 304 and 312 satellites). Finally, the results with the hybrid-simulation mixing both actual data for the MEO observations and simulated one for the LEO are shown demonstrating the feasibility of the methodology. Currently, improvements on the error models are underway expecting to have more realistic simulations for the LEO-PNT components in the future; more receivers, urban scenarios and constellation types are being added, along with new figures of merit to capture better the user needs.