摘要
研究考虑全方位移动机械手系统动力学未知情况下的鲁棒控制问题。首先,在原有动力学模型的基础上,建立考虑外部干扰的移动机械手动力学模型。利用神经网络无穷逼近能力,设计估计器对系统结构不确定性进行在线辨识。然后,提出一种不依赖神经网络先验知识的鲁棒轨迹跟踪控制策略,从理论上证明其稳定性,并且该控制器能够有效阻止非模型有界干扰的影响,实现了对全方位移动机械手系统中不同动力学特性的移动平台和机械臂的协调控制。同时,为了减轻神经网络在线学习的计算量,提出一种分离式的神经网络结构,对系统结构不确定项中的两个独立矢量进行分别辨识,有效地提高了神经网络的训练效率。最后,通过计算机仿真结果验证了所提出控制律的有效性,能够快速稳定地实现全方位移动机械手系统的协调轨迹跟踪控制。
The robust control problem of mobile manipulators with unknown system dynamics is addressed. Firstly, the dynamics of mobile manipulator with external disturbances is constructed based on the primary system dynamics. In virtue of the infinite approximation ability of neural network, an estimator is designed to on-line identify the uncertainties of system structure. Then a neural network robust control scheme without apriori knowledge is proposed. The controller stability is proved in theory and also has the capability of disturbance-rejection in the presence of time varying bounded disturbances. The coordinated control of different dynamics parts： mobile platform and manipulator, is achieved. In addition, in order to reduce the computation burden of the neural network, a partitioned neural network structure is proposed to identify the two separate vectors of the structure uncertainties respectively. The training efficiency of the neural network is highly improved. Finally, computer simulation results validate the effect of the proposed controller, which can cooperatively track the system trajectory quickly and stably.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2009年第1期42-49,共8页
Journal of Mechanical Engineering
基金
国家自然科学基金(60475030
60621001)
智能科学与技术联合实验室基金(JL0605)资助项目。
关键词
全方位移动机械手
神经网络
鲁棒控制
轨迹跟踪
Omnidirectional mobile manipulator Neural network Robust control Trajectory tracking
