Research On Gravity Balanced Upper Limb Rehabilitation Exoskeleton Using Tendon-Sheath Actuator

Stroke is a severe neurological disease caused by occlusion or rupture of blood vessel in brain. It is characterized by the high disease rate, death rate, disability rate and recurrence rate, leading to the hemiplegia and motor dysfunction of patients. In the traditional clinical treatment of hemiplegia, physical therapists have to provide one-on-one treatments to the patients in manual way. The traditional treatments are labour-intensive and high-cost. Besides, the therapeutic efficacy is directly affected by the treatment method, individual experience, subjective judgment, and physical fatigue of physical therapist. Rehabilitation training robot is an important branch of medical robot system integrating many research fields. It can assist or even replace physical therapists to provide patients with sustainable, effective, and multimodal rehabilitation training, and relieve the problem of lacking rehabilitation manpower resources. In addition, rehabilitation training robot helps to strengthen the active training consciousness of patients and, moreover, provide objective evidences to evaluate recovery progress and improve treatment programs and speed up the recovery of motor function. Based on the research background, this dissertation develops a gravity balanced upper limb rehabilitation exoskeleton with tendon-sheath actuator to research the key issues about upper extremity hemiplegia treatment. The main contents of the thesis are presented as follow:Based on the theory of hemiplegia rehabilitation and, moreover, human anatomy and kinematics characteristics of the upper extremity, the design requirements about the applicability, harmony, security, and portability of the upper limb rehabilitation exoskeleton system are proposed. The gravity balanced rehabilitation exoskeleton with tendon-sheath actuation mainly consists of several parts:the shoulder motion module, the elbow and wrist motion module, the shoulder adaptive mobile platform, the tendon-sheath actuation module, and the real-time control system. A semi-physical xPC real-time control platform is established based on Matlab/Simulink/RTW environment. The sensing detection system is also developed based on the platform.The coordinate system and forward kinematics model of the exoskeleton is established. The workspace of the exoskeleton is analyzed by utilizing the Monte Carlo method. The singular configurations of the workspace are specified with triple product method and used to optimize the design of shoulder structure. The Jacobian matrix of the exoskeleton is derived with the vector product method. The redundancy of human upper limb is described via the concept of elbow swivel angle. The "upper limb-mouth coplanar" consumption is proposed to estimate the elbow swivel angle during free movement of upper limb and, moreover, deduce the completed inverse kinematics model. A wearable motion capture system is utilized to analyze the kinematic information of upper limb while conducting four kinds of experiments:drinking water, tuning pages, opening/closing draw, and cleaning desk. The experimental results demonstrate the applicability of the consumption and inverse kinematics model.The gravity balance model of the exoskeleton is established based on the "hybrid balance method". Auxiliary parallel links are used to geometrically locate the center of mass of the overall system. Auxiliary springs, pulleys and cables are utilized to balance the gravity forces and keep the potential energy of system invariant with all available configurations. The balance error model and the robustness of the gravity balanced system are deduced and discussed. The balance error and residual torque of each joint, resulting from the variation of anthropometric parameters, are analyzed by simulation. Further experiments are carried out to compare the surface electromyogram signal of biceps when upper limb perform rehabilitation training with and without gravity balance. The experimental results demonstrate the effectiveness of gravity balance in saving energy.The quasi-static force transmission and elastic extension models of single-tendon-sheath actuation system, as well as the torque transmission model of double-tendon-sheath actuation system are developed based on the Coulomb friction model and force balance of infinitesimal element. An experimental setup for double-tendon-sheath transmission analysis is built to verify the torque transmission model and, moreover, research the effects of system pretension, total curvature of tendon, lubrication condition, and radius of pulley on transmission efficiency. The dynamic model of exoskeleton and upper extremity is deduced with Lagrange method. The friction parameters of robot joints are identified by experiments. The experimental data is analyzed using least square method.Different kinds of rehabilitation training control strategies are developed according to the clinical features of the patients in different therapy periods. A fuzzy sliding mode control algorithm is designed for the stroke patients in acute therapy period, allowing the exoskeleton to provide the patients with passive mode training in predefined trajectories. Based on the minimum interference principle, an admittance control algorithm is designed for the stroke patients in recovery therapy period. It can adjust the assisting forces of exoskeleton according to the motion deviation, helping the patients to perform the assistive mode training. For the patiens in sequela therapy period, an active rehabilitation training strategy is developed based on the admittance control algorithm and virtual reality environment. This strategy can improve the interactivity and interestingness during training. The patients are required to train against the resistance from exoskeleton. The feasibility and effectiveness of each rehabilitation training strategy are experimentally verified.