基于确定学习方法的无人水面艇智能控制研究

项目名称： 基于确定学习方法的无人水面艇智能控制研究

项目编号： No.61473121

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 戴诗陆

作者单位： 华南理工大学

项目金额： 83万元

中文摘要： 在不可预测的动态海洋环境里，无人水面艇通常需要具有良好的自适应、自学能力及快速精确的控制能力来确保自身安全并完成复杂的工作任务。智能是无人水面艇的基本特征和技术难点之一。本项目针对未知海洋动态环境下无人水面艇的轨迹跟踪控制与学习问题，拟提出稳定的自适应神经网络控制方法来实现无人艇控制系统闭环动态的局部准确神经网络辨识（知识获取）、并将所学知识以常值神经网络权值的方式进行表达和存储。在此基础上，拟利用所学知识提出基于动态模式的无人水面艇智能控制方法。该方法将把不同类别的无人艇控制情形定义为不同的动态模式，并对不同的控制情形进行学习和分类；当新的控制情形出现时，采用动态模式估计器来迅速判断是否与曾学习过的控制情形（即经验）相似，并采用切换控制技术选用包含相应知识的神经网络控制器进行控制，实现无人艇快速准确的跟踪控制性能。本项目的研究将为无人艇智能控制理论的发展做出有意义的贡献。

中文关键词： 智能控制；确定学习；自适应控制；神经网络控制；无人水面艇

英文摘要： In the unpredictable dynamic marine environment, unmanned surface vessels (USV) generally need to have an adaptive, self-learning, and accurate-controlling ability for ensuring their own safety and for completing complex tasks. The problem of intelligent control for USV is one of the most important and challenging issues. This project systematically studies the problems of intelligent tracking control and deterministic learning for USV in the unknown marine environment. Stable adaptive neural network (NN) control is presented for uncertain USV. Accurate NN identification of the uncertain USV dynamics can be achieved in the stable control process. The learned knowledge is stored and expressed in a constant radial basis function （RBF） NN. Using the learned knowledge, a pattern-based NN control approach is proposed. Firstly in the training phase, the definitions of dynamical patterns normally occurred in closed-loop control of USV are given. The closed-loop system dynamics corresponding to the dynamical patterns are identified via deterministic learning. In the test phase, sencondly, a pattern classification system is introduced which can rapidly recognize the dynamical patterns in closed loop. If the dynamical pattern for a test control task is recognized as very similar to a previous training pattern, then the NN controller corresponding to the training pattern is selected and activated, which can achieve stability and guaranteed performance of the closed-loop control system without re-adaptation to the controller parameters. The proposed pattern-based NN control approach may provide insight into human's ability to learn and control and possibly lead to smarter robots.

英文关键词： Intelligent control;Deterministic learning;Adaptive control;Nerual network control;Unmanned surface vessels