基于RBF神经网络的机器人滑模控制

RBFNN Mode Sliding Control for Robotic Manipulator

针对不确定机器人系统,把模型分为名义模型和不确定部分两部分,对名义模型可以简化为等效线性系统的标准形式,然后通过极点配置对名义模型进行反馈镇定.针对集中不确定部分,采用径向基神经网络对其进行上界估计,在已估得上界的情况下设计滑模补偿控制器,保证系统的全局稳定,并且利用鲁棒控制项集中补偿有效消除了网络逼近误差,采用饱和函数代替滑模控制中的符号函数,在保证控制效果的前提下有效地消除了控制器抖震现象,利用李亚普诺夫定理证明了控制系统全局稳定,跟踪误差渐近收敛于零。仿真试验结果表明了所提出的控制算法的有效性。

A new control scheme is proposed for robotic manipulators in this paper, it consisted of two combined controllers aim at nominal model and practical plant of the robot system separately. The status feedback controller was designed for nominal model, the compensatory sliding mode controller is designed for practical plant. Radial Basic Function Neural Network （RBFNN） was used to adaptively learn the unknown bounds of system uncertainties, and the robust control focus compensate effectively to eliminate the network approximation error. The stability of the control system was ensured by Lyapunov method. The tracking error converges to zero. In order to make the control smoothed and bounded, a saturated function instead of the symbols function. The simulation studies verify the effectiveness of the proposed algorithm.