【作者】 江先志；

【导师】 熊蔡华； 孙容磊； 熊有伦；

【作者基本信息】 华中科技大学 ， 机械电子工程， 2011， 博士

【摘要】 脑卒中等疾病通常会导致患者肢体运动功能障碍,因为上肢康复机器人能够为患者提供有效的康复运动,促进患者肢体运动功能的恢复,所以社会对上肢康复机器人有日益强大的需求。然而,现有的上肢康复机器人还存在很多不足,如康复运动不够柔顺、安全,或是康复运动类型不够丰富等,因此,本文提出了气动肌肉驱动的外骨骼可穿戴式上肢康复机器人,其各关节(即单自由度的气动肌肉驱动关节)均由两根气动肌肉以双端反向对拉方式驱动,并采取了多种康复运动模式,解决了康复运动不够柔顺、安全和运动类型不够丰富等问题。论文的主要研究内容为：分析了气动肌肉的结构,研究了气动肌肉的静态和动态驱动特性,并对气动肌肉的往返驱动特性作了研究。对气动肌肉驱动关节系统中的张紧装置、钢丝绳与柔性套管间的摩擦力作了分析,研究了气动肌肉驱动关节的单端驱动特性、往返驱动特性,并建立了气动肌肉驱动关节在双端驱动方式下的准静态模型。依据患者在康复期各阶段的不同症状,提出了气动肌肉驱动关节的位置-位置、位置-力矩、力矩-力矩控制模式,其中,前两者主要应用于人体被动-机器人主动的康复运动(简称人体被动运动,或被动运动),力矩-力矩控制模式则主要用于人体主动-机器人被动的康复运动(简称人体主动运动,或主动运动)。针对经典PID控制器在气动肌肉驱动关节控制中的响应慢、超调大等问题,采用神经元PID控制器,在线调节控制器参数,提高了气动肌肉驱动关节的控制性能。由于经典PID与神经元PID在气动肌肉驱动关节的阶跃位置与力矩控制中,很难同时达到响应快与超调小,因此采用模糊控制器,提高了响应速度,抑制了超调。并采用模糊神经网络控制器,进一步提高了气动肌肉驱动关节的响应速度。在气动肌肉驱动关节的轨迹控制中,当气动肌肉驱动关节运动方向改变时,所提出的各类控制器的超调均很大,甚至抖动严重。针对该问题,提出了神经元PID与局部前馈控制结合的复合控制器,有效地抑制了超调,且无抖动。开发了上肢康复机器人控制系统软件,能对康复机器人进行的主动、被动运动控制,同时记录患者的康复运动信息及数据。此外,采用非特定人语音识别技术,实现了康复机器人的语音控制。更多还原

【Abstract】 Disease such as stroke often results in patients with hemiplegia. As upper limb rehabilitation robot can provide patients with effective rehabilitation exercise for recovery, the demand for upper limb rehabilitation robot is increasing nowadays. However, the existing upper limb rehabilitation robots are not compliant and secure enough, or can only provide simple rehabilitation exercises. Therefore, this dissertation proposes an exoskeleton wearable upper limb rehabilitation robot with pneumatic muscle driven, and in which each joint (a single degree of rehabilitation arm) is managed by two pneumatic muscles in antagonistic pairs. The issues that the rehabilitation exercises are not compliant, secure and rich enough are solved. The major works are organized as follows:The structure of pneumatic muscle and it’s static and dynamic characteristics are analyzed. The reciprocating driving characteristics of pneumatic muscle are also studied. The tensioning device and friction between wire rope and flexible pipe of the rehabilitation arm system are analyzed. The driving characteristics and properties of the rehabilitation robot of single-ended driving and double-side are studied.Based on the symptoms in different phases of hemiplegic patients in rehabilitation, the control modes of position-position of the arm, location-torque and torque-torque are proposed. The first two are mainly used in the patient-passive-robot-active rehabilitation exercises (patient passive motion or passive motion for short). The torque-torque control mode is primarily used in patient-active-robot-passive motion (patient active motion or active motion for short) of the rehabilitation robot.For the limitations such as slow response and big overshoot of classic PID controller, neural PID controllers are applied to the control of rehabilitation arm and the control performances are increased, which are benefitted by the advantages of neural PID such as online regulating of its parameters. For classic PID and neural PID, the characteristics of responding quickly and overshoot of small control effect could not be achieved at the same time in control of rehabilitation arm. Fuzzy controllers are applied to position and torque control of rehabilitation arm, which improve the speed of response and almost eliminate overshoot. Furthermore, fuzzy neural network controller achieves more quick response in position control of rehabilitation arm. In tracking control of the rehabilitation arm, the overshoots of all the above controllers are greate when the direction of the rehabilitation arm is changing, and even with serious vibration. To address this problem, a new controller combines neural PID and feedforward control is proposed and the overshoot decreases to a acceptable value without vibration in tracking control of the rehabilitation arm.The control software of upper limb rehabilitation robot is developed, which enables the active and passive motion control of the rehabilitation robot. At the same time, the patient’s rehabilitation information and data are recorded. In addition, speaker independent speech recognition technology is adopted which enables speech control of the rehabilitation robot.更多还原