刚柔耦合系统神经网络跟踪控制方法研究

摘要

针对刚柔耦合系统的姿态跟踪及振动抑制问题,采用一次近似动力学模型,提出了一种神经网络控制方法.典型刚柔耦合系统由一个中心刚体和带有末端质量块的柔性附件组成,假设系统参数未知并且截断模型具有任意的有限维.该方法利用多层神经网络补偿系统非线性项及未建模动态,利用其自学习和自适应能力,降低不确定性因素对系统产生的影响;并且控制器只利用姿态角和角速度信息进行反馈控制,不需要知道柔性附件振动信息.理论分析和实验结果表明该方法使闭环系统所有信号一致最终有界,能有效地使系统完成姿态跟踪,而且能显著地减少柔性结构的弹性振动.%The problem of attitude tracking and vibration suppression for rigid-flexihle coupling systems is dealt with in this paper. A first-order approximation model is applied for the control design. The rigid-flexible coupling system considered consists of a rigid hody and a flexible appendage with a tip mass, and it is assumed that its parameters are unknown and the truncated model has finite but arbitrary dimension. In order to alleviate the effects of nonlinearities and uncertainties, an on line learning multilayer neural network ( MNN ) control strategy is proposed. Only the attitude angle and its derivative are accessible for feedback and elastic modes are not measured. It is shown that the proposed control algorithm guarantees uniform ultimate boundedness of all the closed-loop signals. Physical experiment results also show that the attitude tracking control and vibration suppression can be accomplished effectively.