摘要
研究柔性关节机械手的自适应控制策略,当机械手系统参数准确已知时,传统的反演控制算法可以根据状态反馈对柔性关节机械手进行控制。但是机械手模型参数存在误差时,传统的控制方法需要关节加速度反馈,这将对柔性关节机械手的控制信号将引入噪声,能够破坏系统的动态品质。为解决上述问题,在反演控制算法的基础上引入鲁棒性,提出了鲁棒自适应反演控制算法。在已知模型误差界的条件下,通过神经网络对误差在线自学习,实现了无需模型的柔性关节自适应控制。与传统算法相比,新方法对未知扰动等模型具有鲁棒性及全局稳定性,同时不需要关节加速度信息反馈。
The self-adaptive control scheme has been researched for flexible-joint robots.When the system parameters are known,the backstepping design method is directly applicable to control flexible joint robot manipulators with state feedback.On the other hand,when the system parameters are unknown,in order to control flexible joint robot manipulators,the original adaptive backstepping design method requires the joint acceleration feedback,which are prone to noise.In order to overcome the problem,we present an additional robust control law in conjunction with the adaptive backstepping design procedure.With the known error boundary of model,the self-adaptive control for flexible-joint robots is realized with unknown model parameter.Compared with most of the available control schemes for flexible joint robot system that assumes weak joint flexibility or knowledge of joint accelerations,the proposed control law guarantees global stability of the robot manipulators with uncertain joint flexibility without recourse to any joint acceleration or jerk measurements.
出处
《计算机仿真》
CSCD
北大核心
2011年第2期244-247,共4页
Computer Simulation
