内容：
针对船舶操纵运动建模,本文提出了一种辨识建模方法,即局部最优的局部加权学习算法。该算法通过样本点重新排序和输入空间升维,解决了船舶运动状态一对多映射和不可分问题;并运用留一交叉验证为每个样本点训练一个距离测度,运用加权最小二乘在局部邻域中直接预测船舶操纵运动状态变量。构造局部目标函数,避免了传统的全局最优LWL算法容易陷入局部最优问题。与传统的机理建模相比,局部最优的局部加权学习算法克服了由多重共线性而引起的参数漂移和模型中存在未建模动态问题。通过一组人工数据和3自由度的Mariner轮的学习,实现了对非线性系统的高精度建模。与BPNN预报相比,具有较强的泛化能力。 Aiming at the modeling of ship maneuvering motion, this paper presents a method of identification modeling, that is, local optimal local weighted learning algorithm. The algorithm solves the problem of one-to-many mapping and non-separability of the motion state of the ship by reordering the sample points and increasing the dimension of the input space. By applying a cross validation to train a distance measure for each sample point, the weighted least squares are used to The field predicts the motion state variables of the ship. The local objective function is constructed to avoid the traditional global optimal LWL algorithm which is easy to fall into the local optimal problem. Compared with the traditional mechanism modeling, the local optimal localized learning algorithm overcomes the parameter drift caused by multiple collinearity and the unmodeled dynamic problem in the model. Through a group of artificial data and 3 degrees of freedom of the Mariner wheel learning, to achieve a high-precision modeling of nonlinear systems. Compared with BPNN forecast, it has strong generalization ability.

文件列表（点击上边下载按钮，如果是垃圾文件请在下面评价差评或者投诉）: