PPP-based UAV modeling using Galileo HAS orbit/clock products and RTS smoother

Unmanned aerial vehicle (UAV) remote sensing and mapping using real-time kinematic (RTK) require reference-station corrections and suffer from accuracy degradation over long baselines. To address these limitations, we evaluate the performance of UAV mapping using Precise Point Positioning (PPP) with Galileo High Accuracy Service (HAS) corrections. A Rauch-Tung-Striebel (RTS) smoother is employed to mitigate the convergence issues inherent in PPP. Various PPP solution modes are assessed using dual-frequency L1/L5 UAV observations across multiple platforms and orbit/clock products. Comparisons were conducted from multiple perspectives, including UAV PPP accuracy, checkpoint accuracy, model-derived perimeter, two-dimensional (2D) area, three-dimensional (3D) surface area, and volume. Results show that UAV PPP using rapid products approached RTK, whereas real-time products reduced precision. The horizontal standard deviation (STD) of checkpoints using Galileo HAS products was 6.5 cm, comparable to RTK and to PPP using rapid products; all were less accurate than the post-processed kinematic (PPK) result of 5.3 cm. The root-mean-square (RMS) errors with Galileo HAS were 5.7 cm (North), 12.9 cm (East), and 22.9 cm (Up), indicating pronounced systematic biases attributable to the short observation span and the lower accuracy of real-time products. Regarding model-derived metrics, the perimeter and 2D area from Galileo HAS PPP solutions were comparable to RTK and were minimally affected. By contrast, vertical systematic biases propagated into the 3D metrics: surface-area errors were <0.9% for both rapid and real-time PPP, whereas volume errors were approximately 3% for the rapid PPP solution and approximately 8% for the near real-time PPP solution.