Research On Assist-as-Needed Control Method For Lower Limb Rehabilitation Robot

Rehabilitation robot is an important tool to assist patients in physical therapy and has the great potential to replace conventional therapists,since it can provide repetitive and precise movements and record important information of patient's status.Elderly people are vulnerable to neurologic impairments,such as stroke,and many of these survivals suffer from motor impairments.Taking into account the recovery requirements of these patients,the goal of rehabilitation training is twofold: one is to improve the coordination and dexterity of limbs and the other is to increase muscular strength.Moreover,in the field of robot-assisted rehabilitation,the robot control method should take into account patient's voluntary efforts,since providing too much robotic assistance may result in the slack of patient and negative consequences of rehabilitation training.In this thesis,to encourage the patient's active participation into rehabilitation training,the assist-as-needed(AAN)control method considering patient's pathological characteristics and rehabilitation needs in different recovery stages is studied.The research content in this thesis mainly contains the following parts:For patients in the preliminary stage of rehabilitation,physical therapy should aim at improving the coordination of limbs.Taking the lower limb rehabilitation robot as the experimental platform,impedance model based human-robot interaction control is studied.On this basis,the adjustment strategy of robotic assistance is further discussed and an AAN control method based on adaptive impedance model and virtual tunnel is proposed.The proposed method is capable of adjusting the robotic assistance according to the subject's movement performance and can be applied to improve the coordination and dexterity of limbs.Experiments were performed to evaluate the feasibility and effectiveness of the proposed method.Once the coordination and dexterity of limbs are greatly improved,muscular strengthen training should be performed.Since physical capability can not be quantitatively estimated from the motion measurement,musculoskeletal model is applied to estimate the subject's muscle force and joint torque.To this end,the modeling method of musculoskeletal model is studied and a subject-specific musculoskeletal geometry sub-model is proposed to calculate the muscle-tendon length and muscular moment arm.Furthermore,based on the theories of forward and inverse dynamics,the parameters of musculoskeletal model are identified by off-line optimization methods.Finally,the feasibility and effectiveness of the model are validated and analyzed.The musculoskeletal model based human-robot interface is designed and realized in real-life environment.On this basis,a musculoskeletal model based AAN control method is proposed and achieved by using the impedance model based force control.In order to estimate the requirements of rehabilitation training,the dynamics of lower limb is studied.The proposed control method is capable of quantitatively adjusting the robotic assistance according to the subject's active efforts and can be applied to improve the outcome of muscular strengthen training.