SIA OpenIR  > 水下机器人研究室
水下滑翔机海洋采样方法研究
Alternative TitleMethods Research of Underwater Gliders for Ocean Sampling
朱心科1,2
Department水下机器人技术研究室
Thesis Advisor王晓辉 ; 俞建成
ClassificationTP242
Keyword水下滑翔机 Kriging估计 路径规划 覆盖控制 自适应采样
Call NumberTP242/Z83/2011
Pages104页
Degree Discipline模式识别与智能系统
Degree Name博士
2011-05-30
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract为了理解、建模和预测海洋现象,海洋学家一直试图通过了解某些海洋属性在时间和空间上的分布、变化来解决这一问题。水下滑翔机作为一种新型水下观测平台,具有成本低、作业时间长、航程远,轨迹可控等特点,适合进行大范围、长时间、高分辨率的海洋观测。由于海洋系统是一个动态时变的复杂过程,如何利用当前的观测预报将来的状态以及如何根据当前的预报设计下一步的采样策略,是当前研究的热点,同时依然是一个悬而未决的问题。论文旨在根据水下滑翔机的特点,结合海洋观测实际需求,解决水下滑翔机在海洋监测过程中的实际应用问题,具体包括以下几个方面的研究内容:(1)    水下滑翔机采样优化设计研究。以Kriging估计方差作为海洋采样优化准则,利用贪婪算法得到一组次优的海洋采样点,构建了子模函数,证明了次优解的边界下限。针对大尺度、采样点数较多的海洋采样任务,提出了一步迭代Kriging方差估计和Voronoi图分割采样空间的搜索算法,能够加速算法的搜索速度。仿真实验证实,更新Kriging算法与Voronoi图分割加速算法结合起来使用时,与基本的算法相比,算法速度能够至少提高一个数量级。 (2)    水下滑翔机的使命规划研究。首先,根据水下滑翔机的运动特点,建立了水下滑翔机采样作业过程的能耗模型,提出了一种能耗最小的滑翔运动参数优化方法。针对单个水下滑翔机的使命规划问题,设计了两步链式 Lin-Kernighan算法,能够以较小的时间代价得到更好的采样路径,尤其适合采样点较多的情况。针对多个水下滑翔机的使命规划问题,设计了扩散算法,调节各个采样子空间的采样点,一方面能够保证单个滑翔机的采样路径最优,另一方面使得完成整个采样任务需要的时间最小。最后,以Sea-wing水下滑翔机为例,对上述方法进行了仿真验证。(3)    海流环境中的水下滑翔机路径规划研究。首先,建立了海流模型,推导了在海流影响下滑翔机的运动模型;然后,基于海流预报,我们分别应用Wavefront算法和A*算法进行了路径规划。当出发点和目标点的距离太大,超过了一个海流预报周期时,只能分段进行路径规划,每段最优路径组合得到的完整路径与全局最优路径可能有较大的偏差。针对这个情况,提出了以LCS指导滑翔机路径规划的方法,得到的路径接近于全局最优。(4)    动态海洋特征的覆盖采样研究。首先,定义了基于质心Voronoi分割采样空间的最优采样网络覆盖准则,设计了分布式控制算法,能够保证各个水下滑翔机从任意的初始位置收敛于定义的最优的覆盖采样网络配置。针对海洋特征是完全未知的情况,利用带遗忘因子的递归最小二乘在线估计海洋特征参数,并且证明了参数估计收敛的稳定性。由于海洋特征复杂多变,设计了自适应覆盖采样算法,该算法不仅适用于静态的、动态的海洋特征采样,而且也适用于具有间断性变化的海洋特征。
Other AbstractTo understand, model and forecast the ocean phenomena, oceanographers seek measurements of certain properties across a range of spatial and temporal scales. The underwater glider as a new underwater observing platform, with features of low cost, long endurance, large voyage and controllable trajectory, is suitable to monitor on a large scale ocean phenomena. Marine system is a complicated and dynamic time-varying process. How to use the current state of observation to forecast the future state and how to design the next sampling strategies according current forecast are the current research hot spots and still are unanswered questions. Combining with the glider’s feature and needing of ocean observation, the thesis seeks to solve the practical application problems in which we survey the ocean. The main contents in the thesis include as follows: (1)        The optimal sampling of underwater gliders. The Kriging variance is defined as the performance metric of ocean sampling. Constructing a submodular function, we get the low bound of the sub-optimal sampling set deserved by greedy algorithm. For the large scale sampling task, the one-step recursion Kriging variance estimation and Voronoi partition the sampling space methods are presented, which can improve the algorithm’s search efficiency. Comparing with the basic algorithm, when combining updating Kriging algorithm and Voronoi partition the sampling space, the algorithm’s speed is improved more than one order. (2)        The mission planning of underwater gliders. Firstly, according to the motion feature of underwater glider, we construct the energy consumption model of the underwater glider during ocean sampling course and propose an approach to optimize the gliding parameters. For the mission planning of the single glider, two step chained Lin-Kernighan algorithm is presented aimed at the better sampling path at low cost of time, which suits to numerous samples. For the mission planning of mulitple gliders, we propose a diffusion algorithm to adjust the number of samples in different sampling sub-space, which not only optimize every sampling path but also minimize the time spent on the whole task. Finally, applying the true parameters of Sea-wing underwater glider, methods proposed above are proved by the simulation. (3)        Path planning of the glider in stronger current. Firstly, we contract the ocean current model and the motion model of the glider in current environment. According to current forecast, we use the wavefront algorithm and A* algorithm to get the optimal path respectively. We have to break up the path and optimize every segment when the time spent on the distance between the start and the destination beyond the current forecast period. However, the path combined with all optimal segments is far away from the global optimal path. Applying LCS, a path near to the global optimal path is obtained. (4)        The coverage sampling for dynamic ocean feature. The centroidal Voronoi partition the sampling space is defined as the metric to quantify sampling network configuration. We propose the distributed control law which can guarantee the sampling network to converge to the optimal sampling network configuration from random initial state. Given the unknown ocean feature, the recursive least-squares with forgetting factor algorithm is used to estimate the ocean feature parameters, which is proven asymptotic convergence. Because the ocean feature is complex and changeable, we propose an adaptive coverage sampling algorithm, which is not only suitable for the static and dynamic ocean feature but the non-continuous ocean feature.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/9283
Collection水下机器人研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
朱心科. 水下滑翔机海洋采样方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2011.
Files in This Item:
File Name/Size DocType Version Access License
水下滑翔机海洋采样方法研究.pdf(4755KB) 开放获取LicenseApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[朱心科]'s Articles
Baidu academic
Similar articles in Baidu academic
[朱心科]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[朱心科]'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.