Quadrotor wind estimation: discrete SE(3) model + UKF boosts accuracy

A new study by Hiranya Udagedara, Adam Bigsby, and Mahdis Bisheban (to appear at IEEE CCTA 2026) tackles the challenge of wind velocity estimation using quadrotor UAVs. Instead of the common continuous or discretized dynamic models, the team applied a Lie Group Variational Integrator based on discrete Lagrangian — a true discrete dynamic formulation on the SE(3) manifold. They compared this discrete model against continuous dynamics using both Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF).

The experiments covered numerical simulations (MATLAB) under hover and trajectory-tracking conditions, plus real outdoor free flights. The standout result: pairing the discrete SE(3) dynamics with UKF delivered the highest estimation accuracy, outperforming continuous models even when using low-cost onboard sensors. This suggests that the discrete approach preserves geometric structure better, reducing drift and nonlinearity errors typical of continuous approximations.

The implications extend beyond wind estimation. The same discrete+UKF pipeline could improve state estimation, obstacle avoidance, and precision maneuvers — all while keeping hardware costs down. The paper is available on arXiv (2606.30804) and highlights a practical path to making quadrotors more capable in real-world wind fields without expensive sensor suites.