智能Kalman滤波在水下地形组合导航系统中的应用

摘要

为了避免现实环境的动态变化对水下航行器系统模型造成的随机干扰影响,保证水下航行器长时间导航精度的稳定性,提出了利用RBF神经网络辅助联邦Kalman滤波方法对SINS /TAN/ DVL/ MCP组合导航系统进行信息融合.给出了各子导航系统的误差模型,通过足够精度的样本对前向神经网络进行离线训练,建立神经网络控制模型.仿真结果表明,该方法可使水下航行器的系统状态在较短的时间内以较高的精度达到稳定.通过与联邦Kalman滤波结果对比表明,采用智能控制方法辅助的信息融合方式的导航定位精度提高了一倍,能有效提高常规联邦Kalman滤波器的自适应能力,达到减小误差,提高精度的目的.%In order to avoid the random effects of the underwater vehicle system model caused by the real dynamics of the environment, a RBF neural network aided federated Kalman filter is proposed to fuse the navigation information from various navigation sensors (such as SINS, TAN, DVL and MCP) to improve the navigation accuracy. Error models of navigation systems are proposed and a neural network model is set up via offline training with enough precise samples. Simulation experiments prove that the AUV navigation accuracy can attain high stability in a short period by the proposed method. By comparing the filtering results with the federated Kalman filter, it is shown that the proposed method can effectively improve the adaptivity, and the navigation precision is doubled.