Bio-inspired design and non-linear model predictive control of a self-aligning gait rehabilitation robot

  • Yinan Jin

    Student thesis: Doctoral Thesis


    The field of robot-assisted rehabilitation has seen significant development in recent years. With the development of compliant robots that can be safely used in proximity to people, the use of robots to assist rehabilitation has increased rapidly. The need for gait rehabilitation robots arises from the increasing number of people who are affected by conditions that impair their ability to walk. These conditions can include neurological disorders such as strokes, spinal cord injuries, and traumatic brain injuries. In traditional gait rehabilitation, patients receive manual therapy from a team of physical therapists. While manual therapy can be effective, it can also be time-consuming and resource-intensive, and therapists may not be able to provide consistent and precise support to patients. Gait rehabilitation robots, on the other hand, provide a consistent and precise form of therapy that may help patients make faster and more significant progress. Gait rehabilitation robots can also help reduce the physical demands on therapists and improve the efficiency of therapy sessions. This can allow more patients to receive therapy, which can improve access to care and reduce the burden on health care systems. However, most of existing robotic orthoses have not applied appropriate self-aligning mechanism, gravity-balancing mechanism, or compliant actuators. These limitations should be considered in this proposed research. This thesis proposes a novel intrinsically compliant gait rehabilitation robot with multiple actuated degrees-of-freedom (DOFs). The robot design is flexible and can be personalised with the use of telescopic pelvis, thigh, and shank sections. This newly designed rehabilitation robotic orthosis has multiple actuated and passive DOFs. Because of the importance of alignment between the designed rehabilitation robot joints and human anatomical joints, the robot design has a self-aligning mechanism. A novel gear-couple mechanism, toothed cam-couple mechanism and four-bar linkage mechanism are designed and applied to the hip, knee, and ankle joints to align the robot joints with anatomical joints during gait rehabilitation. Simulation-based and motion capture system-based tests are applied to those three mechanisms to evaluate and choose the most effective self-aligning mechanism. The gear-couple mechanism is finally chosen to be applied to the prototype design. A partial gravity-balancing mechanism is also applied to the designed rehabilitation robot. Gravity-balancing can help overcome the inertia of the rehabilitation robot and can further help reduce joint misalignment. The compliance in the robot is intrinsic due to the use of pneumatic muscle actuators (PMAs). The PMAs have been carefully selected to provide the required torques at the hip, knee and ankle joints during gait rehabilitation. Mechanical amplification of the actuation from the PMAs has been achieved by using gear-couples to replace the usual revolute robot joints. However, with the increase in flexibility of the designed prototype and application of PMAs, which are nonlinear actuators, it is challenging to design the robot control system. This challenge was overcome by developing a system dynamic identification model based on the Koopman operator for the design of a nonlinear model predictive controller (NMPC). The new robot design, together with its self-aligning and gravity-balancing mechanisms, is discussed in detail in this thesis. Compliant actuation and its amplification are explained and various algorithms that are designed and implemented on the robot system as robot firmware are explained. A NMPC is designed and developed to control the rehabilitation robot. The experimental setup and evaluation of the robot design, together with the nonlinear model predictive controller, was carried out with healthy users and yielded the intended results. The robotic orthosis along with the NMPC could successfully guide the healthy human subject along the pre-defined trajectory.
    Date of Award2023
    Original languageEnglish
    SupervisorShahid Hussain (Supervisor), Roland Goecke (Supervisor) & Wayne Spratford (Supervisor)

    Cite this