一种AUV垂直面控制模型的动态辨识算法

Dynamic Identification Algorithm of AUV Vertical Linear Maneuvering Modeling

基于无人自主潜航器(AUV)的实测试验数据,针对AUV多变量、非线性、参数时变的系统特征构建其垂直面控制模型问题,研究讨论了一种面向试验数据的搜索、聚类、模式分类和线性辨识技术相结合的模型辨识算法。首先,利用模型参数的协方差矩阵作为性能指标,对数据点进行初步聚类;然后,通过四步骤调整得到最终的数据点聚类;最后,根据数据点的聚类进行基于线性支持向量机的区间分割和子模型辨识。仿真结果表明,运用这种算法对AUV垂直面控制模型进行的辨识是有效的。

A dynamic identification algorithm of the AUV vertical linear maneuvering modeling based on the date of the AUV test was introduced. The algorithm exploited the combined use of search, clustering, pattern recognition, and liner identification techniques based on the test date. Firstly, a search strategy using the empirical covariance matrix of the model parameters as the evaluation criterion was applied and the rough clusters were obtained. Secondly, a refinement procedure comprising four parts was applied to regulate the initial clusters and the final clusters are decided. Finally, the partition of the system and the parameters identification of each sub-model based on the LSVM could be done easily. The efficiency of the dynamic identification algorithm to the AUV vertical linear maneuvering modeling was demonstrated by simulation examples.