Covariance ellipse-guided base station selection for tightly coupled PPP/INS/5G UAV positioning

Abstract Degraded satellite geometry can lead to anisotropic position uncertainty in PPP/INS solutions. Selecting base stations solely by distance does not explicitly account for the directional characteristics of this uncertainty. To address this limitation, this study proposes a covariance ellipse–guided base station selection strategy for tightly coupled PPP/INS/5G positioning. The strategy uses the major-axis direction of the PPP/INS horizontal covariance ellipse to identify the direction of greatest position uncertainty. Four-station combinations are screened using three-dimensional geometric conditioning and predicted posterior variance constraints, and eligible combinations are ranked by giving priority to variance reduction along this direction. The method is evaluated using terrestrial UAV flight data and semi-simulated 5G ranges in a terrestrial configuration and a simulated nearshore configuration with one-sided base station geometry, each tested with and without simulated GNSS blockage. Across the four cases, the proposed solution reduces the three-dimensional root mean square (RMS) position error by 48.22%–73.94% and the 95th-percentile horizontal position error by 40.75%–74.49% relative to PPP/INS. Compared with nearest-four base station selection, the largest additional reductions are 6.31% in three-dimensional RMS error and 4.80% in the 95th-percentile horizontal error, both obtained in the terrestrial case with simulated GNSS blockage. In the simulated nearshore cases, the additional three-dimensional RMS reductions remain below 0.10%. These results support using directional position uncertainty to guide 5G ranging selection, while showing that the additional gains vary across the tested configurations.