Research On Parallel Robots And The Control Theories And Technologies For Lower Limb & Ankle Rehabilitation

Robot-assisted rehabilitation has recently attracted more and more attentions due to its advantages in terms of high precision,good repeatability,and various training modes over traditional manual therapy.Multiple degrees-of-freedom(DOF)parallel robot is very suitable for lower limb and ankle rehabilitation because of the compact structure,high load and small error capacity.It is also essential to enhance the rehabilitation outcomes by considering the patient's intention and accordingly providing subject-specific cooperative training to emulate the manual therapy.This thesis aims at designing and developing control theories and technologies of multi-DOF parallel robots for lower limb and ankle rehabilitation,and this research was conducted relying on the National Natural Science Foundation of China and the Medical Research Foundation from University of Auckland.The robot mechanism and model,position control,human robot interaction control and cooperative control methods were studied and verified by theories and experiments.In this thesis two multi-DOF parallel robots were designed and implemented for lower limb and ankle rehabilitation respectively.Main contributions of this thesis include:(1)The kinematic and dynamic models of a rigid six-DOF parallel robot were studied in order to estabilish the robot system and control it smoothly for lower limb rehabilitation.A high-performance joint space velocity control method was developed.To overcome the nonlinear uncertainties during human robot operation,a fuzzy sliding mode controller was proposed by combining the advantages of fuzzy system and sliding mode,which will obtain a good comprise between control precision and chattering and it is also verified by the experimental results in real-life.(2)To faciliate active interaction control of the lower limb rehabilitation robot,new virtual tunnel methods based on impedance model were developed,in which the robot trajectories and impedance parameters were adjusted by the human interaction force.As traditional EMG triggered methods are usually divided into two separate parts,this study proposed an adaptive impedance control method to realize a seamless active robot assistance based on motion recognition and muscle evaluation.The robot assistance was triggered by EMG recognition and an impedance controller is used to realize human robot interaction by considering human muscle activity level.Experiments on nine participants verified that the robot assistance is specific to individual patient and it can enhance the training method's adaptability.(3)A soft ankle rehabilitation robot that is redundantly driven by four pneumatic muscle actuators(PMAs)was further designed and developed.The kinematic and dynamic models were studied and the relations between end effector orientation and torque and the joint length and force were investigated.As the muscles must be kept in tension during operation,a force distribution method based on analysis-iterative algorithm was developed for the robot and thus a new control structure including position control loop and force control loop was established.Experiments showed that the robot safety and controllability can be guaranteed by the control system.(4)To improve the robot control performance during repetitive training,a robust normalized iterative feedback tuning(NIFT)controller was proposed and a multiDOF instance was implemented on the parallel robot.This model-free method can achieve a better and better performance iteratively and the experiments with human participants verified that the controller parameters can be optimized rapidly,which ensured good robustness when encountering different rehabilitation situations.(5)To encourage patient's training motivation,a passive-active control method based on periodic trajectory adaptation was proposed.The patient's ankle stiffness was also modelled and used to adjust the robot impedance parameters,so that the robot compliance can be regulated to suit different patients' recovery status.A patientcooperative control strategy was further developed by integrating the proposed active training methods into a virtual reality game.Preliminary clinical trials on patients were conducted on ankle rehabilitation robots to explore the effectiveness of robotassisted rehabilitation.This study validated the feasibility of rehabilitation robots and provides potential support for their future applications in clinical environments.