基于干扰观测器的自适应抗干扰滤波方法及应用

Filtering is the key factors to multi-sensor information fusion performance. Conventional filtering methods include Kalman filter, robust filter and their extensions. Kalman filter is valid only for linear systems with additive Gaussian noises. Robust filter is based on the worst-case design, which may become extremely conservative in some cases. To address the above problems, anti-disturbance filter, where a disturbance observer is adopted to estimate the disturbances whose dynamics are partially known, has been proposed and achieved satisfying results. The existing disturbance observers, however, have assumed that the dynamics of disturbance are precisely known and neglected the fact that in practice, the dynamical model of the disturbance may contain unknown parameters or parameter perturbation. To deal with this issue, we propose in this project an adaptive disturbance observer where the unknown parameters are estimated online. This adaptive disturbance observer will then be combined with the conventional filtering method to give a novel anti-disturbance filter. The adaptive disturbance estimation method will improve the structure of the existing disturbance observers, broaden the application scope of anti-disturbance filters and enrich the anti-disturbance filtering theory. Finally, we will build a semi-physical simulation platform to verify the effectiveness of the proposed method. The expected achievements of this project will refine the anti-disturbance filtering theory and improve both accuracy and robustness of the multi-sensor information fusion system, which is theoretically significant and has great engineering value.

滤波是决定多传感器信息融合性能的关键因素。传统滤波方法有Kalman滤波、鲁棒滤波及它们的推广。Kalman滤波器限定于高斯白噪声系统，鲁棒滤波器基于最坏情况下设计，保守性大。为克服以上问题，基于干扰观测器的抗干扰滤波考虑干扰特性，设计干扰观测器并在此基础上构造抗干扰滤波器，已经取得了一些较好的成果。然而，现有干扰观测器大多假设干扰模型精确，忽略了干扰模型存在未知参数或参数摄动的情形。因此，本项目将针对这类干扰模型不确定性问题，设计一种新型自适应干扰观测器，采用自适应算法在线实时估计未知参数，从而改善已有干扰观测器的结构，扩大干扰观测器的使用范围。然后与传统滤波方法结合，设计一种新型自适应抗干扰滤波器，应用于组合导航系统。最后搭建半物理仿真实验平台验证方法的有效性。本项目的研究成果将进一步完善干扰观测及抗干扰滤波理论，提高组合导航系统的精度和抗干扰能力，具有重要的理论意义和潜在的应用价值。

滤波是决定多传感器信息融合性能的关键因素。传统滤波方法有Kalman滤波、鲁棒滤波及它们的推广。Kalman滤波器限定于高斯白噪声系统，鲁棒滤波器基于最坏情况下设计，保守性大。为克服以上问题，本课题研究基于干扰观测器的自适应抗干扰滤波方法，并应用于组合导航系统验证其有效性。本课题按照计划顺利完成以下研究工作：.1）针对惯导/卫星/偏振/地磁组合导航系统，考虑三大干扰源：外部干扰、量测误差和建模误差，建立了包含多源干扰的惯导/卫星/偏振/地磁组合导航系统模型。.2）针对一类存在未知外部系统参数矩阵的非线性系统，构造状态观测器和自适应干扰观测器估计出系统状态和外部干扰；并设计基于自适应干扰观测器的输出反馈复合控制器。.3）针对受未知干扰影响的系统，研究同时估计未知干扰和线性系统状态的问题。在没有未知干扰先验知识的前提下，针对系统方程中未知干扰系数矩阵不满秩的情况，分别针对带直通项和不带直通项的系统，设计两种新型抗干扰滤波方法。.4）完成了组合导航系统的软硬件设计以及各个导航子模块的标定、测试，并完成了四旋翼飞行器硬件测试平台搭建和调试工作。.5）针对偏振/地磁辅助的组合导航系统性能进行了一系列实验验证。验证主要分为两个部分，第一部分针对的是惯导/卫星/偏振/地磁组合导航系统，第二部分针对的是惯导/偏振/地磁组合导航系统。.除此之外，为了以四旋翼飞行器为硬件平台测试组合导航系统，设计了一系列基于干扰观测器的控制器，保证四旋翼飞行器在干扰影响下仍能完成相关飞行任务。还将自适应干扰观测器的设计思想应用于航天器的姿态控制系统，做了一系列算法研究和仿真，验证自适应干扰观测器估计干扰的效果。.本项目的研究成果将进一步完善干扰观测及抗干扰滤波理论，提高组合导航系统的精度和抗干扰能力，具有重要的理论意义和潜在的应用价值。