An Intelligent Pneumatic Muscle Actuated Exoskeleton for Robotic Gait Rehabilitation

Abstract

Gait disorder is a commonly lasting side-effect for stroke and spinal cord injury survivors. Conventional gait rehabilitation trainings provided by therapists are largely dependent on their experience. Such trainings are often challenging for the therapists due to their physically intensive nature. Hence, consistent optimal results cannot always be achieved. Robotic technologies were thus introduced to automate the gait rehabilitation trainings, in order to emancipate therapists from physically intensive work as well as making rehabilitation training more accessible to patients Research have shown that task specific repetitive training and patients’ active participation can lead to more effective gait rehabilitation. However, conventional trajectory tracking controlled robotic gait rehabilitation could change the dynamics of the walking task, reduce inputs from patients’ motor systems, lower their physical effort and thus result less effective outcomes. Therefore, it is important to ensure that the robotic gait rehabilitation training is more analogous to actual human walking and maximize the training subject's active participation. The goal of this thesis is the development of a new robotic GAit Rehabilitation EXoskeleton (GAREX) that is compliant with the current neurorehabilitation theories in order to achieve optimised robotic gait rehabilitation. Such goal is tackled systematically in terms of both robotic design and control algorithm research. GAREX was designed to provide safe, task specific gait rehabilitation to stroke patients. Pneumatic muscles (PM) actuators were used to drive GAREX, due to their high power/force to weight ratio and intrinsic compliance. Specially, the intrinsic compliance can create a wide range of dynamic environment for control strategy development. However, the negative correlation between PM’s force output and contracting length means a trade-off between torque and range of motion specifications of the actuation system. The design of GAREX comprehensively addressed torque and joint range of motion requirements imposed by task-specific gait rehabilitation training. Control strategies are the key to implement the training theories into robotic operations. In order to encourage patients' active participation, the robot should be controlled to supply just enough guidance/assistance a patient needs to complete treadmill based gait training. To implement assist-as-needed (AAN) concept, the robot should also be able to assess the extent of active participation and change the assistance provided accordingly. The intrinsic compliance of GAREX’s PM actuation system could be utilized to change the level of guidance. A new multi-input-multi-output (MIMO) sliding model (SM) controller was developed to adjust assistance while guiding training subjects to walk in predefined gait trajectories. Technical experimental validation indicated that controller was able to track reference gait trajectories and the desired joint space average antagonistic PM pressures. A study with 12 healthy subjects revealed strong statistical evidence that the proposed MIMO SM controller is able to vary the compliance of the exoskeleton To online assess the training patient’s active participation, a fuzzy logic compliance adaptation (FLCA) controller is proposed. The FLCA algorithm utilizes the robotic kinematics and human- exoskeleton interaction torque of the knee joint, to estimate the extent of the patient’s active participation. Based on the estimation, the desired compliance level can be automatically adjusted with higher compliance for more active participation and vice versa. Nevertheless, the FLCA algorithm does not require models of the exoskeleton and biomechanics of the training subject, which means less preparation work and easier implementation. Performance of the FLCA control system was validated with three healthy subjects who simulated different extents of participation. The FLCA control system could successfully adapt the joint actuation compliance accordingly in all the scenarios.