摘 要

当前世纪，由于曾经制约智能船舶发展的各项技术瓶颈被突破，智能船舶的发展趋势日益迅猛。智能船舶相较于传统水面舰船具有作业更高效、任务更标准、人力资源消耗更少等优点。因而为了抢抓海洋发展机遇，促进船舶行业改革，我国制定了《智能船舶发展行动计划（2019-2021年）》。在该行动计划中明确强调要重点开展传感器数据融合技术研究，研制船舶状态等数据采集与数据融合的感知系统。由此可见对智能船舶感知系统进行研究具有重要意义。

在获取智能船舶状态信息时，针对单种传感器无法满足多种工况需要的问题，智能船舶通常会配备多种传感器来获取数据。由于传感器的自身特性不尽相同，多种传感器的测量数据并非指向同一参考条件下的船舶状态信息。并且，单传感器的测量数据会因自身故障、外界干扰等因素发生异常跳变乃至丢失的现象。最后，传感器通常受到噪声干扰以及自身测量误差的影响，致使其获取得到的数据通常包含噪声等干扰项，并不能真实反映船舶实际运动状态。为获取准确的船舶状态信息，为智能船舶控制模块提供准确可靠的信号输入，本文针对上述问题，以GPS和INS两种传感器数据为基础，围绕智能船舶数据融合以及状态估计这一主题，开展了如下工作：

（1）针对传感器自身特性不同而无法进行数据融合的问题，本文分别采用SPL（Space Precision Location）法、自适应野值剔除法以及自适应外推法对传感器数据进行了空间校准、野值剔除以及时间校准。并通过仿真实验验证了算法的有效性。

（2）针对单传感器无法满足智能船舶航行过程中多种工况需求，且测量数据容易受到外界干扰等问题，本文以两种传感器为基础，采用自适应加权融合算法以及一种基于改进权值动态分配算法的自适应加权融合算法分别对预处理后的两种传感器数据进行数据融合。并通过仿真实验证明了所构建算法理论的可行性，得出改进的自适应加权融合算法有效地提高了融合精度。

（3）针对传感器测量数据通常会包含噪声等干扰项的问题，分析传统卡尔曼滤波仅可用于线性系统这一局限性，本文基于扩展卡尔曼滤波算法进行智能船舶状态估计。对智能船舶状态空间模型进行Taylor展开，并保留一阶线性项，以达到模型线性化的目的。最终完成状态估计理论流程构建，并通过仿真实验证明了基于扩展卡尔曼滤波算法进行状态估计可有效地获得智能船舶低频运动估算信息。

关键词：智能船舶；数据融合；状态估计；自适应加权融合；扩展卡尔曼滤波

Abstract

During the present century, due to the breakthrough of various technical bottlenecks that had restricted the development of intelligent ships, the development trend of intelligent ships is increasingly rapid. Compared with traditional surface ships, intelligent ships have the advantages of more efficient operation, more standard tasks and less consumption of human resources. Therefore, in order to grab the chance of marine progress and facilitate the innovation of the shipping industry, our country formulated "intelligent ship development action plan (2019-2021)”. In this plan, it is clearly emphasized to focus on the research of sensor data fusion technology, and to develop a sensing system for data collection and data fusion, such as ship status. Thus, it is important to investigate the intelligent ship sensing system.

When acquiring the status information of the intelligent ship, the intelligent ship is usually equipped with a variety of sensors to obtain the data for the problem that a single type of sensor cannot meet the needs of multiple working conditions. Due to the different characteristics of sensors, the measurement of various sensors do not point to the ship status information under the same reference condition. In addition, the measurement of a single sensor may jump or even be lost due to its failure, external interference and other factors. Finally, the sensor is usually affected by noise interference and its own measurement error, so the acquired data usually contain noise and other interference items, which cannot truly reflect the actual motion state of the ship. In order to obtain accurate ship status information and provide accurate and reliable signal input for the intelligent ship control module, this thesis, based on GPS and INS sensor data, focused on the topic of intelligent ship data fusion and state estimation, carried out the following work:

(1) To solve the problem of data fusion due to the different characteristics of sensors, this thesis adopts SPL (Space Precision Location) method, adaptive wild value elimination method and adaptive extrapolation method to carry out spatial calibration, wild value elimination and time calibration of sensor data. Simulation results show the effectiveness of the algorithm.

(2) In view of the single sensor cannot meet the demand of intelligent ship sailing in the process of a variety of working conditions, and the measurement data are susceptible to interference and other issues, this article is based on two kinds of sensors, uses the adaptive weighted fusion algorithm and a dynamic allocation algorithm based on improved weight adaptive weighted fusion algorithm respectively to data preprocessing of two kinds of sensor data fusion. The feasibility of the proposed algorithm is proved by simulation experiments, and the improved adaptive weighted fusion algorithm can effectively improve the fusion accuracy.

(3) Aiming at the problem that sensor measurement data usually contain noise and other interference items, this thesis analyzes the limitation that traditional Kalman filter can only be used for linear systems. Based on the Extended Kalman filter algorithm, this thesis conducts intelligent ship state estimation. Taylor expansion is carried out on the intelligent ship state space model, and the first-order linear term is retained to achieve the purpose of model linearization. Finally, the theoretical process of state estimation is constructed, and the consequence of simulation indicate that the state estimation based on the Extended Kalman filter algorithm can effectively obtain the low-frequency motion estimation information of the intelligent ship.

Key Words: Intelligent ship; Data fusion; State estimation; Adaptive weighted fusion; Extended Kalman filter

目 录

第1章 绪论 1

1.1 研究目的及意义 1

1.2 国内外研究现状 2

1.2.1 数据融合 2

1.2.2 状态估计 4

1.3 本文研究内容及章节安排 5

第2章 智能船舶运动数学模型 8

2.1 引言 8

2.2 坐标系及参数定义 8

2.3 船舶运动数学模型 9

2.3.1 六自由度运动学模型 9

2.3.2 六自由度动力学模型 10

2.4 智能船舶数学模型 12

2.4.1 船舶低频运动模型 12

2.4.2 船舶高频运动模型 13

2.4.3 环境力模型 14

2.4.4 传感器测量模型 14

2.4.5 非线性运动数学模型 14

2.5 本章小结 15

第3章 多传感器数据融合 16

3.1引言 16

3.2传感器及介绍 16

3.2.1 GPS工作原理 16

3.2.2 INS工作原理 17

3.3传感器数据预处理 19

3.3.1 空间校准 20

3.3.2 野值剔除 21

3.3.3 时间校准 22

3.3.4 仿真结果与分析 23

3.4传感器数据融合 26

3.4.1 自适应加权融合 26

3.4.2 改进自适应加权融合 28

3.4.3 仿真结果与分析 30

3.5 本章小结 31