长春工业大学学报

基于INS/GNSS/DR融合导航系统，提出了一种自适应联邦卡尔曼滤波算法，通过融合多传感器数据并自适应调整权重，提升了复杂环境下的导航精度和可靠性。首先构建了两级结构的联邦卡尔曼滤波模型，实现对无人机位置、速度和姿态的高精度估计。在此基础上，提出基于测量噪声协方差的自适应权重调整机制，用于动态分配子系统的信息权重。仿真实验结果表明，当DR的导航精度下降时，自适应联邦卡尔曼滤波器仍能有效跟踪真实轨迹，位置误差收敛在±2 m以内，速度误差小于1.5 m/s,相比传统联邦滤波，位置精度提高了约40%,收敛速度加快了50%。

This article proposes an adaptive federated Kalman filtering algorithm based on INS/GNSS/DR fusion navigation system, which improves navigation accuracy and reliability in complex environments by fusing multi-sensor data and adaptively adjusting weights. Firstly, a two-level federated Kalman filter model was constructed to achieve high-precision estimation of drone position, velocity, and attitude. On this basis, an adaptive weight adjustment mechanism based on measurement noise covariance is proposed for dynamically allocating information weights to subsystems. The simulation experiment results show that when the navigation accuracy of DR decreases, the adaptive federated Kalman filter can still effectively track the real trajectory, with position error converging within ± 2 m and velocity error less than 1.5 m/s. Compared with traditional federated filtering, the position accuracy is improved by about 40%, and the convergence speed is accelerated by 50%.