基于快速故障检测的AUV组合导航容错策略研究

The autonomous underwater vehicle (AUV) has become a powerful assistant for human to explore and develop the ocean because of its safety, high efficiency and long underwater operation. The navigation system provides position, speed and attitude information for the AUV. Accurate and reliable navigation system is one of the guarantees for the AUV to successfully perform its tasks. This project is devoted to the research of fault-tolerance strategy for AUV integrated navigation systems, aiming at improving the reliability of it. Firstly, breaking through the assumption that the reference system is fault-free, a self-detection module is set up for the strapdown inertial navigation system. A novel fault detection function is designed to improve the sensitivity of the navigation system to gradual faults. Secondly, aiming at the unstable distribution of the measurement noise in underwater environments, an adaptive filtering algorithm based on a variable length sliding window is proposed to improve the stability of the filter. Different processing methods are adopted to deal with navigation subsystems with different lengths of failure time. Finally, an information distribution strategy based on the probability of failure is designed to enhance the anti-pollution ability of fault-free subsystems in AUV integrated nabvigation systems. The research can effectively improve the reliability and fault-tolerance of AUV integrated navigation systems, thus providing a powerful guarantee for the safe navigation of AUVs.

自主水下航行器(Autonomous Underwater Vehicle, AUV)以其安全、高效，可长时间水下作业的特征，成为人类探索和开发海洋的有力助手。导航系统为AUV提供位置、速度、姿态信息，准确可靠的导航系统是AUV成功执行任务的保证之一。本项目致力于研究AUV组合导航系统的容错策略，旨在提高其导航系统的可靠性。首先，突破参考系统无故障的假设，建立捷联惯性导航系统故障自检测模块，构造新型故障检测函数增强导航系统对渐变故障的敏感度；其次，针对水下工作环境中系统量测噪声分布不稳定的问题，提出基于非定长滑动窗的自适应滤波算法以提高滤波稳定性，并采用不同的应对方式处理失效时长不同的导航子系统；最后，设计基于故障概率的信息分配策略，增强AUV组合导航系统中无故障子系统的抗污染能力。本项目的研究可有效提高AUV组合导航系统的可靠性、容错性，为AUV的安全航行提供有力保障。

自主水下航行器(Autonomous Underwater Vehicle, AUV)作为探查海洋资源、开展深海研究的重要工具，在我国海洋事业发展中发挥重要作用。AUV导航系统为其提供位置、速度、姿态信息，导航结果的准确性和可靠性直接关系到AUV能否顺利完成任务并安全返航。本项目针对AUV组合导航系统故障检测技术、多传感器信息融合技术、容错处理技术展开研究，取得了一系列成果。建立了AUV组合导航系统容错结构，设计故障检测模块和故障隔离模块。提出了基于多输出支持向量回归的故障检测方法，所设计的新型故障检测函数对渐变故障具有较强的敏感度，车载导航实验证明该方法与残差卡方故障检测法相比具有更快的检测速度。针对复杂水下环境造成的量测模型不稳定情况，设计了模糊自适应滤波算法，实现AUV导航系统量测噪声方差阵的在线调整。针对特殊工况下导航子系统发生短时失效的情况，提出了基于强高斯过程回归的量测估计器，改进的高斯过程回归算法针对样本集进行多次训练形成数个弱估计器，弱估计器与模型权重相结合构成强估计器，实验证明该方法可有效处理长达150秒的量测中断，有效提高了导航结果精度。提出了基于量测可信度的分段故障隔离策略，利用新息变化率和故障检测函数值联合确定量测可信度，充分利用了渐变故障临界区域的有效信息，从而提高了导航结果的稳定性。针对AUV各导航子系统量测更新频率不一致的问题，设计了基于多尺度的异步信息融合算法，提高了融合结果精度。本项目在AUV组合导航系统故障检测和容错处理方面取得了实质性进展，为实现高精度、高可靠性、高容错性的AUV组合导航系统提供了有效方法和依据。