Research On Pneumatic Muscles Driven Wrist Rehabilitation Robot And Lower-limb Exoskeleton And Their Control Technology

Based on the theoretical principle of neuroplasticity,rehabilitation training can let the patients who are suffering from limb motor dysfunction regain their daily activities.Rehabilitation robots can assist the patients to perform repeatable and precisely training exercise,so that it can reduce the burden of therapists,increase the training time and training strength and improve the effect of rehabilitation.Meanwhile,the rehabilitation robots can evaluate patients' recovery status by analysing the data recorded in robotic training process.The study of rehabilitation robot is still in the initial stage in our country.Therefore,it is urgent to develop the rehabilitation robots with our independent intellectual property rights.In this paper,with the aim of safety and low cost,2 kinds of rehabilitation robots,which are used for wrist joint rehabilitation and gait orthosis,are developed by using pneumatic artificial muscle actuators.From the point of view of safe control,a new type of proxy-based sliding mode control method has been introduced as the position control for the developed robots' joints.Aiming at the problem of the lack of robustness in the position control of the rehabilitation robots driven by PAM actuators by using PSMC,it has been improved by using a single neuron and neural network respectively to self tune the local parameters of the PSMC.The improved PSMC has better robustness without changing the characteristic of safety.Firstly,to understand the output characteristics of pneumatic muscles and provide the basis for rehabilitation robot design,we improves the existing pneumatic muscle model.Based the fact that the thickness and elastic modulus of the rubber cylinder change with the change of the PAM length,the elastic force term in the static model of PAM is modified.The mathematical expression of the correction coefficient of elastic force is obtained by experiments.The accuracy of the PAM model has been improved.Secondly,as the existing wrist joint rehabilitation robot can only be used for the patients with grasping ablility,and seldom takes the compatibility of the upper-limb rehabilitation robot and the hand exoskeleton into account,a kind of 2-DOF wrist rehabilitation robot which is driven by pneumatic artificial muscles(PAMs)and suitable for the patients who lose the ability of grasping is developed in this paper.It has the advantages of reducing the bearing load of the patient,avoiding the interference with the finger movement,the natural process of the movement,and the convenience and comfort of wearing.Thirdly,to solve the problems of low quality,high cost and poor flexibility of the existing lower limb exoskeleton rehabilitation robot,a new type of lower-limb exoskeleton rehabilitation robot driven by PAMs is developed in this paper.It can drive the human hip and knee joint to perform the flexion and extension movement and realize the gait rehabilitation training.It has the advantages of small weight,low cost and good safety.Fourthly,as most of the existing low-level position control of rehabilitation robots aim at improving the accuracy and ignore the safety,we combine the proxy-based sliding mode controller and the independent joint compliance controller to improve the safety of the joint control.PSMC controller can generate a slow,damping and smooth resume motion from the unexpected large position errors as the strong safety of the control,and can also ensure the tracking accuracy to the small position errors.Based on the relationship between the safety and the internal force and the relationship between the internal force/compliance and the nominal pressure of a PAM actuator,the independent joint compliance controller(IJCC)can reduce the inner force without influencing the position control to improve the safety performance of the PAM actuator.At last,in the prototype experiments,the PSMC method with fixed parameters is found to lack robustness for the load change when it is used as the position controller of a PAM actuator.Thus,this paper proposes 2 methods to improve PSMC controller:using single-neuron-PID controller instead of the PID controller in PSMC or introducing BPNN to adjust the equivalent PID coefficient in the saturation function of PSMC control law.By the online learning ability of a neuron or neural network,the parameters of PID can be online self-tuning.By these ways the robustness of the PSMC is improved without changing the safety performance.The effectiveness and feasibility of the improved parameter-self-tuning PSMC as the joint position controller is verified by the experiment of the robot prototypes worn by human subjects.