The accurate noise parameter is essential for the Kalman filter to obtain optimal estimates. However, problems such as variations in the noise environment and measurement anomalies can cause degradation of estimation accuracy or even divergence. The adaptive Kalman filter can simultaneously estimate state and noise parameters, while its performance will also be degraded in complex noise. To address the problem of estimation accuracy degradation and result divergence of the integrated navigation system in a complex time-varying noise environment, an improved multiple-model adaptive estimation (MMAE) that combines the Sage–Husa adaptive unscented Kalman filter with the MMAE is proposed in this paper. The forgetting factor is included as an unknown parameter of MMAE so that the algorithm can adjust the value of the forgetting factor according to different system states. In addition, we improve the hypothesis testing algorithm of classical MMAE to deal with the competition problem of undesirable models that severely impacts the performance of variable-parameter MMAE and enhance the algorithm’s parameter identification capability. Simulation results show that this method enhances the system’s robustness to noises of different statistical properties and improves the estimation accuracy of the filter in time-varying noise environments.

中文翻译：

---

时变噪声综合导航的改进多模型自适应估计方法

准确的噪声参数对于卡尔曼滤波器获得最佳估计至关重要。然而，诸如噪声环境的变化和测量异常等问题会导致估计精度下降甚至发散。自适应卡尔曼滤波器可以同时估计状态和噪声参数，但在复杂噪声下其性能也会下降。针对复杂时变噪声环境下组合导航系统估计精度下降和结果发散的问题，提出了一种将Sage-Husa自适应无迹卡尔曼滤波器与MMAE相结合的改进多模型自适应估计（MMAE）。在本文中。遗忘因子作为 MMAE 的未知参数包含在内，使得算法可以根据不同的系统状态调整遗忘因子的值。此外，我们改进了经典MMAE的假设检验算法，以解决严重影响变参数MMAE性能的不良模型的竞争问题，增强算法的参数识别能力。仿真结果表明，该方法增强了系统对不同统计特性噪声的鲁棒性，提高了滤波器在时变噪声环境下的估计精度。我们改进了经典MMAE的假设检验算法，以解决严重影响变参数MMAE性能的不良模型的竞争问题，增强算法的参数识别能力。仿真结果表明，该方法增强了系统对不同统计特性噪声的鲁棒性，提高了滤波器在时变噪声环境下的估计精度。我们改进了经典MMAE的假设检验算法，以解决严重影响变参数MMAE性能的不良模型的竞争问题，增强算法的参数识别能力。仿真结果表明，该方法增强了系统对不同统计特性噪声的鲁棒性，提高了滤波器在时变噪声环境下的估计精度。