SIA OpenIR  > 水下机器人研究室
AUV 水下对接智能归航方法研究
Alternative TitleResearch on Intelligent Homing Method in AUV Docking Application
董凌艳1,2
Department水下机器人研究室
Thesis Advisor封锡盛
Keyword水下对接 导航 导引
Pages137页
Degree Discipline机械电子工程
Degree Name博士
2020-12-06
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract自主水下机器人(Autonomous Underwater Vehicle, AUV)因其作业范围更广、自主性和隐蔽性更高而被广泛应用。然而AUV受其自身携带的能源和存储空间的限制无法实现水下长期作业。AUV在海洋环境下的布放和回收成本高,并且作业人员伴随着一定的风险性。近年来,AUV的重点发展方向是实现水下长期驻留作业。水下长期驻留作业无需频繁地布放和回收AUV,在节约成本的同时,增加了AUV的自主性和隐蔽性。对于一些需要长航程和长航时的科考任务或者大深度的下潜任务,AUV水下长期驻留作业有助于收集更多的观测数据。当AUV结束使命或日常维护保养时,AUV的自主回收能够节约成本并且降低作业人员的风险性。AUV水下自主对接技术是实现AUV水下长期驻留作业和自主回收的关键。 水下对接基站通常呈现两种状态,一种是保持位置和朝向不变的固定状态,另一种是以某种规律运动的移动状态。对应地,AUV和固定基站及移动基站的对接分别称为静态对接和动态对接。AUV水下自主对接过程包含导航和导引两项关键技术。导航算法的目的是使AUV精确估计其自身的位置和对接基站的位置。导引算法的目的是使AUV按照规划路径或行为归航至对接基站。导航算法的性能影响对接的成功率和效率,导引算法的性能影响归航过程的安全性及能源消耗。海洋环境的复杂性使AUV携带的导航传感器的测量结果带有高斯噪声或非高斯噪声,并且噪声的统计特性是未知或不准确的。传感器的未知或不确定噪声会导致导航性能的下降,进而导致对接成功率和效率的下降。为了减小传感器噪声的影响,通常的解决方案是导航算法以滤波算法为基础,但滤波算法对噪声统计特性的准确性高度敏感,未知或不准确的噪声统计特性会降低滤波的精度,甚至导致滤波发散。在导航传感器存在测量噪声条件下,为了使AUV的静态对接和动态对接具有较高的对接成功率和效率,并且归航过程满足安全性及低能耗的要求,本文的研究内容如下:(1) 针对导航传感器的不确定高斯噪声影响AUV静态对接过程中的导航精度问题,提出一种鲁棒的导航算法。采用基于非线性滤波的FastSLAM2.0算法实现静态对接归航过程中的AUV和固定基站的同时定位。由于导航传感器中用于测速的速度传感器和用于测量航向角的姿态传感器的测量噪声统计特性具有不确定性,不准确的噪声统计特性会导致FastSLAM算法的导航精度下降。在此提出采用模糊Q学习算法改进FastSLAM算法。改进算法通过和环境的交互作用增强算法的鲁棒性,使算法在噪声统计特性不准确时,仍然具有较高的导航精度。MATLAB仿真和试验数据分析结果证明改进后的算法具有更优的导航性能。(2) 针对导航传感器的未知非高斯噪声影响AUV动态对接过程中的导航精度问题,提出一种自适应导航算法。采用基于扩展卡尔曼滤波的目标跟踪算法实现移动基站运动状态未知时AUV和移动基站的同时定位。由于导航传感器中用于测速的速度传感器和用于测量航向角的姿态传感器的测量噪声统计特性未知,并且移动基站运动状态存在估计误差,对此提出采用基于变分贝叶斯算法估计导航噪声协方差及移动基站运动状态的估计误差协方差。为了减小导航噪声及移动基站运动状态的估计误差导致的系统状态估计误差,在线训练一个基于神经网络的误差补偿器。MATLAB仿真和实验数据分析结果表明,动态对接过程中导航传感器存在未知非高斯噪声时,自适应导航算法具有很好的导航性能。(3) 针对超短基线定位系统的未知非高斯测量误差影响AUV静态对接和动态对接归航过程中的导航精度问题,提出离线算法和在线算法相结合的导航方法。根据超短基线的误差特点和AUV的运动模型,提出归航过程中的观测误差数据的收集方法。采用基于高斯混合模型的变分自编码算法估计观测误差整体分布的统计特性。采用支持向量回归算法拟合工作距离和误差统计特性之间的非线性关系。基于离线算法估计的观测噪声统计特性用于在线算法对状态估计的更新阶段。实验数据分析结果表明在水声传感器的观测误差统计特性未知时,所提算法可以有效估计观测误差的统计特性,并减小观测误差对导航性能的影响。(4) 针对AUV自主对接归航过程中的导引问题,分别提出适用于静态对接和动态对接的归航导引方法。AUV静态对接归航过程应该满足平稳性、安全性及低能耗的要求,在此提出基于Dubins曲线理论的最优路径规划方法。采用视线法实现对规划路径的跟踪。在AUV动态对接归航过程中,为了获得移动基站状态的最大似然估计,提出基于费雪信息矩阵增量行列式最大的导引策略。MATLAB仿真验证了导引策略的有效性。(5) 采用MATLAB仿真验证基于本文所提导航算法和导引算法的AUV静态对接和动态对接的归航结果。通过对比不同条件下算法的对接成功率和效率,验证本文所提算法的有效性。便携式AUV的湖上航行试验证明了静态对接方法的有效性。为了验证导航算法的性能,大量应用了由中国科学院沈阳自动化研究所海洋信息技术装备中心自主研发的AUV和回收系统的湖上试验数据。最后详细介绍了试验所用的载体和试验装置。
Other AbstractAutonomous Underwater Vehicle (AUV) is widely used because of its wider operating range, higher autonomy and concealment. However, due to the limitation of energy and storage space, AUV cannot realize long-term underwater operation. The cost of AUV deployment and recovery in marine environment is high, and the operators are accompanied with certain risks. In recent years, the key development direction of AUV is to realize long-term underwater residence operation. There is no need to deploy and recover AUV frequently for long-term underwater residence operation, which can save cost and increase the autonomy and concealment of AUV. For some scientific research missions that need long voyage mileage and sailing time or deep diving missions, the long-term underwater residence operation of AUV is helpful to collect more observation data. When the AUV ends its mission or routine maintenance, the autonomous recovery of AUV can save cost and reduce the risk of operators. AUV underwater autonomous docking technology is the key to realize the long-term operation and autonomous recovery of AUV. The underwater docking station usually presents two states, one is the fixed state which keeps the position and orientation unchanged, and the other is the mobile state which moves with certain regularity. Correspondingly, the docking of AUV with fixed docking station and mobile docking station is called static docking and dynamic docking respectively. The underwater autonomous docking process of AUV includes two key technologies: navigation and guidance. The purpose of the navigation algorithm is to enable the AUV to accurately estimate its own position and the position of the docking station. The purpose of the guidance algorithm is to make AUV return to the docking station in a certain path or behavior. The performance of navigation algorithm affects the success rate and efficiency of docking, and the performance of guidance algorithm affects the safety and energy consumption of the homing process. Due to the complexity of the marine environment, the measurement results of navigation sensors carried by AUV have Gaussian noise or non Gaussian noise, and the statistical characteristics of noise are unknown or inaccurate. The unknown or uncertain noise of sensors will lead to the degradation of navigation performance, which will lead to the decline of docking success rate and efficiency. In order to reduce the influence of sensor noise, the usual solution is that navigation algorithm is based on filtering algorithm, but the filtering algorithm is highly sensitive to the accuracy of noise statistical characteristics. Unknown or inaccurate statistical characteristics of noise will reduce the accuracy of filtering, and even cause filtering divergence. In order to make the static docking and dynamic docking of AUV have high success rate and efficiency, and meet the requirements of safety and low energy consumption in the homing process, the research contents of this paper are as follows: (1) Aiming at the problem that the uncertain Gaussian noise of navigation sensors affects the navigation accuracy of AUV in static docking process, a robust navigation algorithm is proposed. The FastSLAM2.0 algorithm based on nonlinear filtering is used to realize the simultaneous positioning of AUV and fixed docking station (FDS) in the homing process of static docking. Due to the uncertainty of the measurement noise statistical characteristics of the velocity sensor and attitude sensor in the navigation sensor, the navigation accuracy of FastSLAM algorithm will be reduced. In this paper, a fuzzy Q-learning algorithm is proposed to improve FastSLAM algorithm. The improved algorithm enhances the robustness of the algorithm through the interaction with the environment, so that the algorithm still has high navigation accuracy when the statistical characteristics of noise are not accurate. The results of Matlab simulation and Experimental data analysis show that the improved algorithm has better navigation performance. (2) Aiming at the problem that the unknown non-Gaussian noise of navigation sensors affects the navigation accuracy of AUV in the homing process of dynamic docking, an adaptive navigation algorithm is proposed. The target tracking algorithm based on extended Kalman filter (TT-EKF) is used to realize the simultaneous positioning of AUV and mobile docking station (MDS) when the motion state of MDS is unknown. Since the statistical characteristics of the measurement noise of the velocity sensor and the attitude sensor in the navigation sensor are unknown, and there are estimation errors in the motion state of the MDS, this paper proposes to estimate the navigation noise covariance and the estimation error covariance of the MDS motion state based on the variational Bayes algorithm. In order to reduce the system state estimation error caused by navigation noise and motion state estimation error of MDS, an error compensator based on neural network is trained online. The results of MATLAB simulation and Experimental data analysis show that the adaptive navigation algorithm has good navigation performance when there is unknown non Gaussian noise in the homing process of dynamic docking. (3) Aiming at the problem that the unknown non Gaussian observation error of ultra short baseline (USBL) positioning system affects the navigation accuracy of AUV in homing process of static docking and dynamic docking, a navigation method combining off-line algorithm with on-line algorithm is proposed. According to the error characteristics of USBL and the motion model of AUV, the collection method of observation error data in homing process is proposed. A variational auto-encoding (VAE) algorithm based on Gaussian mixture model is used to estimate the statistical characteristics of the overall distribution of observation errors. Support vector regression (SVR) algorithm is used to fit the nonlinear relationship between working distance and error statistical characteristics. The statistical characteristics of observation error based on off-line algorithm estimation are used in the update step of online algorithm. The results of Experimental data analysis show that the proposed algorithm can effectively estimate the statistical characteristics of observation errors and reduce the impact of observation errors on navigation performance when the statistical characteristics of observation errors of underwater acoustic sensors are unknown. (4) Aiming at the guidance problem in homing process of AUV autonomous docking, the homing guidance methods suitable for static docking and dynamic docking are proposed respectively. The homing process of AUV static docking should meet the requirements of stability, security and low energy consumption. The line-of-sight method is used to track the planning path. In the homing process of AUV dynamic docking, in order to obtain the maximum likelihood estimation of MDS state, a guidance strategy based on the maximum incremental determinant of Fisher information matrix is proposed. The results of Matlab simulation indicate that the effectiveness of the guidance strategy. (5) Matlab simulation is used to verify the homing results of AUV static docking and dynamic docking based on the proposed navigation algorithm and guidance algorithm under different conditions. By comparing the success rate and efficiency of the algorithm under different conditions, the effectiveness of the proposed algorithm is verified. The effectiveness of the static docking method is proved by the navigation experiment of the portable AUV on the lake. In order to verify the performance of navigation algorithm, a large number of experimental data of AUV and recovery system independently developed by Shenyang Institute of automation, Chinese Academy of sciences are applied. Finally, the vessel and device used in the test are introduced in detail.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/27984
Collection水下机器人研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
董凌艳. AUV 水下对接智能归航方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2020.
Files in This Item:
File Name/Size DocType Version Access License
AUV 水下对接智能归航方法研究.pdf(26140KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[董凌艳]'s Articles
Baidu academic
Similar articles in Baidu academic
[董凌艳]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[董凌艳]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.