基于支持向量机的船舶航行路径规划

发布时间：2018-02-23 06:52

本文关键词： 路径规划支持向量机多分类问题分段线性函数　出处：《大连海事大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着海上运输和海洋活动的频繁,提高船舶驾驶自动化水平成为重要且迫切需要的课题。船舶航行路径规划是船舶安全航行及自主航行的基础和前提,有着重要的理论和应用价值。本文研究了一种基于多分类问题支持向量机的路径规划方法,进行了理论分析和实验验证。首先,总结了常用的路径规划方法,以及目前支持向量机理论在路径规划问题中的应用。其次,对路径规划问题这一优化问题和C-支持向量机解决模式分类问题的方法进行了描述。通过比较分析,说明了用支持向量机进行路径规划的可行性和需要解决的问题。再次,从最优化理论的角度详细介绍求解C-支持向量机的凸二次规划问题的对偶问题的方法。通过核函数技巧,使得支持向量机能够解决线性不可分问题。本文还提出了一种利用约束Deluanay三角剖分的分段线性支持向量机方法,并与核函数方法搭配,完成每个二分类支持向量机的计算。然后,介绍了支持向量机解决多分类问题的方法,选取了适合路径规划的一对一方法。其中每个二分类机的参数考虑路径光滑性和到障碍物距离的指标单独选取。最后,用C++语言在Mircrosoft Visual Studio 2015集成开发环境中结合CGAL和LIBSVM软件包,对本文选用的方法进行了实验验证。包括环境模型、数据预处理、参数选取、计算路径等过程,得到了距离障碍物较远且较为光滑的规划路径。
[Abstract]:With the frequency of maritime transportation and ocean activities, it is an important and urgent task to improve the level of ship navigation automation. Ship routing planning is the basis and premise of safe and autonomous navigation. This paper studies a path planning method based on multi-classification support vector machine, and carries out theoretical analysis and experimental verification. Firstly, the common path planning methods are summarized. And the application of support vector machine theory in path planning problem. Secondly, the optimization problem of path planning problem and the method of solving pattern classification problem based on C- support vector machine are described. The feasibility of path planning using support vector machine and the problems to be solved are explained. Thirdly, the method of solving the dual problem of convex quadratic programming problem of C- support vector machine is introduced in detail from the angle of optimization theory. The support vector machine can solve the problem of linear inseparability. A piecewise linear support vector machine method based on constrained Deluanay triangulation is proposed and matched with the kernel function method. The computation of each two-classification support vector machine is completed. Then, the method of solving the multi-classification problem is introduced. A one-to-one method suitable for path planning is selected, in which the parameters of each two classifier take into account the smoothness of the path and the distance to the obstacle. Finally, the CGAL and LIBSVM software packages are combined with C language in the Mircrosoft Visual Studio 2015 integrated development environment. The methods used in this paper are verified by experiments, including environmental model, data preprocessing, parameter selection, calculation path and so on.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U664.82;TP18

【相似文献】