智能假肢感知和约束变增益迭代学习控制  

Apperceiving and Variable Gain Iterative Learning Control with Limited Input for Intelligent Prosthesis

作　　者：王斌锐[1] 金英连[1] 许宏[1] 

机构地区：[1]中国计量学院控制理论与控制工程研究所,浙江杭州310018

出　　处：《控制工程》2008年第6期711-714,共4页Control Engineering of China

基　　金：国家自然科学基金资助项目(50475107)

摘　　要：针对膝关节输入力矩为受限的阻尼力矩,步态跟随控制缺乏在线调节问题,提出用离线和在线集成的变增益一阶D型闭环迭代学习方法进行步态跟随控制。建立系统动力学模型,给出基于有限状态机的步态感知和人腿摆动相步速间接感知方法,减小跟随滞后。设计在线变步长指数增益调节算子,并针对输入力矩受限,设计离线的带惩罚因子的指数增益调节算子。利用谱半径约束条件,分析算法的收敛性。在真人比例虚拟样机模型上进行控制仿真实验,给出了固定增益和变增益调节下轨迹跟踪曲线。结果证明,提出的增益调节算子可保证输入在有效范围内,并可提高跟踪速度。To the problem that control input of prosthesis knee is limited torque, and amputee gait is varying and affected, the variable gain Dtype closed-loop iterative learning control is proposed and designed for prosthesis gait tracking. System dynamical model is established, and the appereeiving method of gait phase and speed is proposed to reduce tracking delay based on finite state machine. The method of on-line adjusting learning gain is given. Penalty factor is used to correct the off-line varying exponential order learning gain. The convergence is analyzed based on spectral radius. Control simulations on virtual prototype show that the proposed gain varying method can restrict control input within given bound and reduce tracking time.

关 键 词：双足机器人 智能假肢 建模 感知 迭代学习控制 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]