基于进化动态递归模糊神经网络的上肢康复机器人自适应阻抗控制

Adaptive impedance control based on evolutionary dynamic recurrent fuzzy neural network for upper-limb rehabilitation robots

针对机器人辅助患肢进行康复训练时患肢病情的变化对系统运动平滑性和稳定性造成的影响,在传统阻抗控制方法的基础上,提出了一种基于进化动态递归模糊神经网络(EDRFNN)的新的自适应阻抗控制方法。该方法根据在线辨识得到的患肢机械阻抗参数,运用EDRFNN对目标阻抗控制参数进行动态调整。在调整过程中,首先采用混合进化算法离线优化目标阻抗控制参数,然后再利用基于Lyapunov函数稳定收敛性理论设计的动态BP算法对目标阻抗控制参数在线作进一步的调整。分析和仿真结果表明,这种新的方法较其它阻抗控制方法更能有效地适应患肢病情的变化,且具有较好的平滑性和稳定性。

In consideration of the fact that when an upper-limb robot-aided rehabilitation system works the change of the impaired limb＇s physical condition often has influence on the system＇s rehabilitation training efficiency, the paper studies the existing impedance control methods for upper-limb rehabilitation robots, and on the basis of this a new adaptive impedance control strategy based on evolutionary dynamic recurrent fuzzy neural network （EDRFNN） is proposed. The strategy uses an on-line identification method to estimate the impaired limb＇s mechanical impedance parameters, and uses the EDRFNN to dynamically regulate the target impedance control parameters. In the regulating, the hybrid evolutionary algorithm is applied to offline optimize desired impedance control parameters, and then the dynamic back-propagation （BP） algorithm designed based on the Lyapunov theory is used to further on-line adjust the target impedance control parameters. The analysis and simulation results indicate that the proposed algorithm is much more stable and robust than other impedance control methods.