Adaptive Unscented Kalman Filter with Sampling Correction for Trajectory Prediction

Beam tracking based on trajectory prediction can effectively mitigate beam misalignment for highly maneuverable unmanned aerial vehicles (UAVs), thereby enhancing communication link quality and stability. To improve trajectory prediction accuracy and enable real-time error correction, this paper proposes an Adaptive Unscented Kalman Filter with sampling correction (AUKF-SC). The algorithm dynamically adjusts the covariance matrix to balance the weights between state estimation and observation update. Furthermore, a sampling correction strategy is introduced to ensure that sampling points better capture the true state distribution in high-dimensional systems. Simulation results demonstrate that the AUKF-SC algorithm achieves approximately 17.4% higher prediction accuracy than the conventional Unscented Kalman Filter (UKF), while maintaining low computational complexity. Particularly, the proposed method exhibits robust performance during high-speed UAV maneuvers.