Adaptive UKF multisensor fusion localization algorithm for UAV formation

Multi-sensor data fusion technology has improved the positioning accuracy and system stability of multi-UAV formations in complex environments, and has become a mainstream solution. This study integrates data from UWB and Intel RealSense T265 sensors to provide precise positioning information for multi-UAV formations. An improved Unscented Kalman Filter (UKF) algorithm is proposed, incorporating adaptive sigma point generation and a dual-layer adaptive covariance adjustment mechanism. This modified algorithm dynamically adjusts the weight of each sensor in the positioning estimation based on the noise characteristics of different sensors, thereby effectively mitigating the impact of high-noise data on positioning accuracy and significantly enhancing overall positioning performance. Experimental results indicate that the improved UKF algorithm, compared to the conventional UKF, enhances positioning accuracy in the x, y, and z directions by 6.78%, 3.29%, and 2.88% respectively in RMSE evaluation, with corresponding reductions in standard deviations by 4.84%, 1.55%, and 7.04%. Particularly in the z direction, the interquartile range (IQR) improvement is notable, with an increase of 7.60%.