鲍春雷, 王斌锐, 金英连, 柯海森. 气动肌肉关节的无模型自适应CMAC迟滞补偿控制[J]. 机器人, 2015, 37(3): 298-303,309. DOI: 10.13973/j.cnki.robot.2015.0298
引用本文: 鲍春雷, 王斌锐, 金英连, 柯海森. 气动肌肉关节的无模型自适应CMAC迟滞补偿控制[J]. 机器人, 2015, 37(3): 298-303,309. DOI: 10.13973/j.cnki.robot.2015.0298
BAO Chunlei, WANG Binrui, JIN Yinglian, KE Haisen. Model Free Adaptive CMAC Hysteresis Compensation Control of the Pneumatic Muscle Joint[J]. ROBOT, 2015, 37(3): 298-303,309. DOI: 10.13973/j.cnki.robot.2015.0298
Citation: BAO Chunlei, WANG Binrui, JIN Yinglian, KE Haisen. Model Free Adaptive CMAC Hysteresis Compensation Control of the Pneumatic Muscle Joint[J]. ROBOT, 2015, 37(3): 298-303,309. DOI: 10.13973/j.cnki.robot.2015.0298

气动肌肉关节的无模型自适应CMAC迟滞补偿控制

Model Free Adaptive CMAC Hysteresis Compensation Control of the Pneumatic Muscle Joint

  • 摘要: 为补偿迟滞性对气动肌肉关节轨迹跟踪控制精度的破坏,首先建立了关节模型;推导得到迟滞力方程,测试分析了关节迟滞性;而后设计了无模型自适应CMAC 神经网络迟滞补偿算法,该算法采用了充、放气双重结构;采用梯度下降法实时反馈调整充、放气过程的网络权值;采用4 阶傅里叶拟合函数对网络权值降噪;基于高斯函数和邻域误差,设计误差可信度评估函数来调节学习率,抑制干扰对神经网络的影响;而后用三角波轨迹跟踪控制对神经网络进行了学习训练;最后将训练好的神经网络用于突发干扰下的正弦波轨迹跟踪控制.实验结果表明,该算法能自适应非线性曲线跟踪控制中的迟滞变化,有效抑制突发干扰,提高控制精度.

     

    Abstract: In order to compensate the destruction of hysteresis on trajectory tracking control accuracy of pneumatic muscle joint, a joint model is established. Furthermore, a hysteresis force equation is derived accordingly and the hysteresis characteristics are tested and analyzed. A model free adaptive CMAC (cerebellar model articulation controller) hysteresis compensation control algorithm is presented, which adopts the double structure of inflation and deflation. Gradient descent is employed to adjust network weights of the inflation and deflation processes. The fourth order Fourier fitting function is used to denoise network weights. Based on the Gaussian function and the neighborhood error, an error credibility evaluation function is designed to adjust the learning rate and suppress the impact of burst interference on the neural network. Then the triangular wave trajectory tracking control is applied to training the neural network. Finally, the trained neural network is applied to sine wave trajectory tracking control with sudden disturbance. Experimental results show that the control algorithm is self-adaptive to the hysteresis variation in nonlinear curves tracking control. Using this algorithm, not only the sudden disturbance is restrained effectually, but also the control accuracy is improved.

     

/

返回文章
返回