Design analysis and assist-as-needed control of a Stephenson III Six-Bar linkage-based robotic gait rehabilitation orthosis

Abstract

Repetitive and task-oriented movements can strengthen muscles and improve walking capabilities among patients experiencing gait impairments due to neurological disorders. The demand for effective rehabilitation is high, given the large number of patients suffering from gait impairments. The traditional physiotherapy is laborious, may not provide the desired cadence and gait patterns, and requires constant presence of physiotherapists. This often leads to delayed treatment for many patients due to the high demand and a shortage of physiotherapists. Early phase post-stroke gait rehabilitation is crucial, as the ability to recuperate lost muscular abilities reduces over time. Lower limb wearable rehabilitation robots have shown promise in improving the locomotor capabilities of patients experiencing gait impairments and reducing the burden on physiotherapists. However, the high cost of commercially available robots makes this technology inaccessible to many hospitals and rehabilitation centers. To address this issue, ongoing research is focusing on improving existing rehabilitation robots in terms of ease of use, innovative design, and cost reduction. Closed-loop linkage mechanisms have recently drawn attention in the development of gait rehabilitation robots due to their ability to address the drawbacks of commercially available robot orthoses. These mechanisms are affordable and capable of providing suitable trajectories for gait training therapy. One of the challenging aspects in designing linkage-based robots is determining and calculating linkage parameters that will produce the required gait trajectories. This thesis presents an innovative approach to synthesizing the linkage dimensions to provide natural gait trajectories. Additionally, it introduces a novel and affordable robotic orthosis based on Stephenson III's six-bar linkage. The developed gait rehabilitation orthosis is a bilateral system powered by a single actuator on each side of the leg, capable of providing naturalistic knee and ankle joint motions relative to the hip joint, which are required during therapeutic gait training. This orthosis can be used in clinical settings and is actuated using only a single motor, yet it is capable of providing complex lower limb trajectory motions at its end-effector. The initial design optimization was carried out using a genetic algorithm (GA), and a deep generative neural network model was developed for the linkage synthesis problem. This model represents an advancement in current kinematic synthesis methods, enabling it to generate dimensions of the links that satisfy various required target human lower limb trajectories during walking in a short period. It will assist designers in determining optimal linkage dimensions to generate the required end-effector trajectories within a single mechanism. To enhance the mechanism's velocity regulation control scheme and address fluctuations that may occur during operation due to external disturbances such as fixed patient’s leg and inertia in closed loop linkage mechanisms, a Deep Reinforcement Learning control scheme was proposed to regulate the speed of the input crank to reach satisfactory performance needed for gait rehabilitation training. Experimental evaluations with healthy human subjects were conducted to demonstrate that the mechanism is capable of directing lower limbs on naturalistic gait trajectories with a required walking speed. Furthermore, given the varied disability levels among neurologically impaired patients, the orthosis incorporates a patient cooperative control strategy. This is achieved through the application of impedance learning control, operating on an "assist-as-needed" principle. This innovative approach enables the robot to modify the assistive force it provides during gait cycle aligning with the patient’s disability level and contributing towards active participation during the gait rehabilitation training. The proposed control scheme was evaluated in two distinct gait training modes while being worn by a human subject. In the "passive" mode subjects refrained from moving their legs, allowing the robot to guide their movements. While during the second ‘active’ mode, the subject engaged in normal walking activity while wearing the robot. Experimental results with healthy human subjects indicated reduced robot torques consequent to an increase in human torque. These results substantiate that customized robotic assistance based on the individual needs of patients can enhance their participation, which is essential to improve the treatment outcomes. The concept of this research lies in the development of a novel, affordable, and adaptable robotic orthosis based on Stephenson III’s six-bar linkage mechanism, capable of delivering naturalistic individualized lower limb motion. It advances the fields of dimensional synthesis of closed loop linkage mechanisms rehabilitation robotics with the use of deep generative neural network and a Deep Reinforcement Learning control scheme for enhanced velocity regulation. Moreover, the application of impedance learning control encourages active patient participation in gait rehabilitation training by customizing assistive force based on the patient's disability level. With these advancements, the research contributes significantly to the development of more cost-effective, adaptable, and efficient robotic gait rehabilitation systems, presenting a promising solution for improving therapeutic outcomes for patients with gait impairments due to neurological disorders.

Date of Award	2023
Original language	English
Supervisor	Shahid HUSSAIN (Supervisor), Roland Goecke (Supervisor) & Nick Brown (Supervisor)