UAV Navigation with Starlink

The first ever experimental demonstration of unmanned aerial vehicle (UAV) navigation with Starlink low Earth orbit (LEO) satellite signals of opportunity is presented. The navigation system tightly couples Doppler measurements extracted from Starlink downlink signals with an inertial measurement unit (IMU). The satellites’ unknown downlink signals are processed through a cognitive software-defined receiver to extract Doppler frequency measurements. To account for the satellites’ inaccurate publicly available ephemerides, a ground segment (reference receiver with a known position) estimates via the Doppler measurements fused in an unscented Kalman filter (UKF) each satellite’s dynamic states, namely its position and velocity, and communicates ephemeris messages to the UAV. The UAV navigates by fusing Starlink Doppler measurements with IMU data via an extended Kalman filter (EKF) in a tightly coupled fashion along with loosely coupled altimeter measurements. Experimental results are presented showing a UAV navigating with 9 Starlink satellites. The UAV traversed a total trajectory of 500 m in 75 seconds. The unaided IMU navigation solution resulted in a three-dimensional (3-D) position root mean squared error (RMSE) of 41.81 m and a final error of 97.33 m. A comparative study was conducted where the LEO-aided IMU navigation filter was implemented with four different ephemerides products: (i) general perturbations (GP) elements propagated with SGP4, (ii) supplemental GP (SupGP) elements, (iii) SpaceX-published ephemeris, and (iv) ground segment ephemeris. Both types of GP elements were propagated with simplified general perturbations (SGP4), while the third type is already in the form of a satellite position and velocity time series. The 3-D position RMSEs achieved with (i)-(iv) were 102, 22.1, 15.57, and 8.52 m, respectively, with final errors of 217, 23.89, 16.34, and 6.13 m, respectively.