外界扰动下基于多尺度估计理论的智能水下机器人组合导航方法研究

In order to compensate for the defect of single navigation method ,now the AUV integrated navigation system generally consists of sound, light, electricity and other sensors. Due to its special underwater work environment, propagation rate of acoustic sensor ,an important of AUV navigation equipment is low. As a result, the AUV underwater integrated navigation sensor resolution, frequency data are not identical, have significant multiscale effect and multiscale features..This project intends to adopt the method of multiscale estimation theory, several methods for AUV navigation research. This method, combining characteristics of the AUV underwater work environment and the movement contribution to underwater navigation sensor data, sets up the AUV navigation dynamic multiscale model and discusses the AUV underwater integrated navigation system multirate, multiscale characteristics and mechanism . This method researches the AUV underwater integrated navigation method based on multiscale fusion estimation theory and takes the underwater integrated navigation fault diagnosis and error analysis, and carries out simulation, pool and outfield experiments. Through this project research, we are expected to reveal the AUV underwater integrated navigation method based on multiscale estimation theory under current disturbance in theory, verify its feasibility and carry out the theory and technology for AUV to improve the navigation accuracy in various tasks .

为了弥补单一导航方式的缺陷，现在智能水下机器人（AUV）组合导航系统一般由声、光、电等多种传感器构成，由于其特殊的水下工作环境，作为AUV重要的导航设备-声学传感器传播速率低，因此，AUV水下各种组合导航传感器分辨率、数据频率一般不相同，存在显著的多尺度效应与多尺度特性。本项目拟采用多尺度估计理论的方法，进行AUV组合导航数据处理方法研究。结合AUV水下工作环境和运动特点对水下导航传感器数据的影响，建立AUV导航动态多尺度模型，探讨 AUV水下组合导航系统多速率、多尺度特性与机理，研究基于多尺度估计理论的AUV水下组合导航融合方法，进行水下组合导航的故障诊断与误差分析，并开展仿真、水池与外场实验研究。通过本项目研究，可望在理论上揭示海流扰动下基于多尺度估计理论的AUV水下组合导航方法，验证其可行性，为AUV在水下执行各种作业任务提高导航精度奠定理论与技术基础。

本课题针对水下环境中量测噪声统计特性未知或变化的特点，基于多尺度估计理论对外界扰动下AUV组合导航数据处理方法及故障诊断方法进行了研究，主要研究内容包括：建立了一类水下导航系统多尺度模型，提出解决导航系统状态方程非线性问题的提出两种多尺度非线性滤波算法；研究扰动下基于多尺度的异步多速率多传感器数据处理方法，提出了基于小波变换的异步多速率多传感器数据融合估计方法。开展了基于多尺度分析理论的AUV组合导航系统中传感器故障诊断方法研究，提出了一种基于多尺度熵的传感器故障特征提取方法，对现有的小波神经网络做出优化与改进，提出了一种结合了带动量-自适应学习速率算法与Levenberg-Marquardt算法的小波神经网络作为传感器故障分类器；研究水下导航故障或传感器数据测量不完全等条件下基于多尺度理论的故障诊断方法；研究了组合导航系统各种误差产生的原因，通过传感器采样频率等多尺度特性对水下组合导航系统的精度进行补偿，减小了导航误差。