基于极限学习机的机械臂自适应神经控制

Adaptive Neural Control of Manipulators Based on Extreme Learning Machine

针对刚性机械臂系统的控制问题,提出基于极限学习机(ELM)的自适应神经控制算法.极限学习机随机选择单隐层前馈神经网络(SLFN)的隐层节点及其参数,仅调整其网络的输出权值,以极快的学习速度获得良好的推广性.采用李亚普诺夫综合法,使所提出的ELM控制器通过输出权值的自适应调整能够逼近系统的模型不确定性部分,从而保证整个闭环控制系统的稳定性.将该自适应神经控制器应用于2自由度平面机械臂控制中,并与现有的径向基函数(RBF)神经网络自适应控制算法进行比较.实验结果表明,在同等条件下,ELM控制器具有良好的跟踪控制性能,表明了所提出控制算法的有效性.

We propose an adaptive neural control algorithm for the rigid manipulators system based on extreme learning machine （ELM）. The ELM for a single-hidden layer feedforward neural network （SLFN） can analytically determine the output weights of the SLFN and randomly choose hidden nodes and its parameters, providing good generalized performance at an extremely fast learning speed. Using the Lyapunov synthesis approach, the proposed ELM controller can approximate the model uncertainy of systems by adaptively tuning the output weight to guarantees the stability of the overall closed-loop control system. The proposed adaptive neural controller is applied to control a planar manipulator with two degrees of freedom and is compared with the existing radial basis function neural control algorithms. Experiment results show that the ELM controller has good tracking performance at the same experiment conditions, which demonstrates the effectiveness of the proposed control algrorithm.