Journal of Mechanical Transmission
RBF基于 神经网络的闭链下肢康复机器人自适应补偿控制
李东琦1,2 秦建军1,2 孙茂琳1,2 郑皓冉1,2 李 伟3
摘要 在下肢康复机器人的康复训练过程中,模型参数、环境干扰等不确定性因素会影响机器Radial Basis Function, RBF)人轨迹跟踪的精度。针对这一问题,提出了一种基于径向基函数( 神经4网络的自适应补偿控制,该控制方法能够提高机械系统轨迹跟踪的精确性。首先,设计一款具有种工作模式、运动稳定的闭链卧式下肢康复机器人结构;然后,利用拉格朗日方法求解动力学名义RBF模型,将康复装置的模型参数以及外界干扰等不确定性因素分离出来,并设计基于 神经网络的Matlab/Simulink自适应补偿算法对其进行逼近控制;最后,通过 环境对其进行仿真验证,证明了该RBF控制策略的有效性。结果显示,在人体步态曲线轨迹跟踪中,提出的基于 神经网络的自适应补- - Proportional Integral Derivative, PID)偿算法相比传统的模糊比例 积分 微分( 控制的方法响应速度
0. 08° 0. 13°,快、跟踪效果好,且髋关节和膝关节轨迹跟踪的角度误差峰值分别为 和 远小于患者下RBF肢在康复运动中的转动角度。设计了单腿样机试验,试验结果表明,采用的 补偿自适应控制器能够实现高精度的跟踪结果,也能够满足患者在康复训练中安全性的要求。RBF
关键词 下肢康复机器人 闭链结构 神经网络 不确定性 自适应补偿控制Li Dongqi1,2 Qin Jianjun1,2 Sun Maolin1,2 Zheng Haoran1,2 Li Wei3 (1 School of Mechanical-electronic and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 100044, China) (2 Beijing Engineering Research Center of Monitoring for Construction Safety, Beijing 100044, China)
(3 Institute of Sports Medicine, General Administration of Sport of China, Beijing 100061, China)
Abstract In the rehabilitation training process of lower limb rehabilitation robots, the existence of uncer⁃ tain factors such as model parameters and environmental interference will affect the accuracy of trajectory track⁃ ing of the robot. To solve this problem, an adaptive compensation control based on the radial basis function (RBF) neural network is proposed. This control method can improve the accuracy of mechanical system trajecto⁃ ry tracking. Firstly, a closed chain horizontal lower limb rehabilitation robot structure with four working modes and stable movement is designed. Secondly, the Lagrange method is used to solve the kinetic nominal model, the uncertainty factors such as model parameters and external interference of the rehabilitation device are separated, and the adaptive compensation algorithm based on the RBF neural network is designed for the approximate con⁃ trol. Finally, the Matlab/Simulink environment is used to verify the effectiveness of the control strategy. The re⁃ sults show that, compared with the traditional fuzzy proportional integral derivative (PID) control method, the adaptive compensation algorithm based on the RBF neural network has a faster response speed and better track⁃ ing effect in human gait curve trajectory tracking. Moreover, the peak angle errors of the hip joint and the knee joint trajectory tracking are 0.08° and 0.13° respectively, which are much less than the rotation angle of pa⁃ tients' lower limbs in rehabilitation exercise. A single-leg prototype experiment is designed to show that the
Adaptive Compensation Control of Closed-chain Lower Limb Rehabilitation Robots Based on the RBF Neural Network