A study of a Gough-Stewart platform-based manipulator for applications in biomechanical testing.

Abstract

This thesis investigates the development and application of a robotic system for in-vitro biomechanical testing to study the mechanisms leading to human joint injury and degeneration in an ethical and safe manner. Six degree of freedom (6-DOF) robotic-based systems, in particular Gough-Stewart platform-based systems, have been increasingly used in applications of biomechanical testing where 6-DOF mobility, large load capacity, and high stiffness and positioning accuracy are required from the testing machine. This study proposes a novel Gough-Stewart platform-based manipulator with ultra-high stiffness and accuracy for use in biomechanical testing and investigates its mechanism and control. Not only restricted to biomechanical testing, the proposed manipulator concept can also be applied to other robotic-based applications, particularly those requiring ultra-high accuracy positioning under large external loads (e.g machining). Four main features of the proposed manipulator are individually studied in this thesis: namely, stiffness and control of a non-collocated actuator-sensor mechanism, active preload control using actuation redundancy for backlash elimination, adaptive velocity-based load control of human joints for unconstrained testing, and reproducing the in-vivo measured kinematics on human cadaveric joints. Stiffness and Control of the Non-collocated Actuator-Sensor Mechanism A novel Gough-Stewart platform-based mechanism is proposed with a fully decoupled actuator-sensor arrangement for passively compensating the structural compliance of the manipulator. The stiffness of the robot load frame and the sensing frame are respectively quantified using the robot kinematics error model combined with finite element analysis (FEA) on the top and bottom assemblies. Numerical results demonstrate that the proposed mechanism improves the stiffness of the robotic testing system in excess of an order of magnitude on the translational axes and two orders of magnitude for rotational axes compared to a traditional actuator-sensor collocated design. Control disturbances arising from actuator-sensor non-collocation is addressed using decoupled control. Experimental results show that the proposed decoupled control algorithm improves the dynamic accuracy of the manipulator by approximately 25% on average. Active Preload Control Using Actuation Redundancy for Backlash Elimination This thesis investigates combining the benefits of both active and passive preload control methods, using actuation redundancy to prevent backlash on a general Gough- Stewart platform. Both the mechanical configuration and the dynamics model of the redundant manipulator are investigated for the ease of control. A novel online optimization algorithm combined with a feedback force control scheme is formulated to achieve a real-time method which is robust to both model inaccuracy and load disturbance. Simulation results demonstrate an effective preload efficacy by the redundant arrangement within the workspace of the robot. Simulation results also show that the proposed method can effectively achieve backlash-free positioning of the manipulator under large 6-DOF external loads. Experimental results further prove that the proposed method can eliminate backlash instabilities from control and consequently higher bandwidth control can be achieved by the robot with improved accuracy. Adaptive Velocity-based Load Control of Human Joint for Unconstrained Testing A novel adaptive velocity-based load control method is proposed in this thesis to more effectively achieve pure force or moments on human joints under unconstrained testing compared to existing methods. The force/moment control gains are designed to vary adaptively based on the tracking performance of the force/moment to make a compromise between load following and control stability, which makes the proposed method self-adaptive to unknown joint dynamics. Sheep functional spinal units are used to experimentally valida