Task-Space Control of Powered Lower-Limb Prostheses

The field of lower-limb prosthetics is currently in an exciting period of development, with improvements in the design and performance of robotic prostheses making the goal of replicating the capabilities of biological limbs more of a reality. Recent advancements in control algorithms for these devices have achieved joint-level behavior comparable to normative kinematics and kinetics from sensing isolated to the prosthesis itself. However, this isolated sensing causes the user and the device to be treated as separate entities, where the prosthesis must infer the state of the user. Research in human motor control suggests a more integrated coupling, where joint-level responses arise as the result of, rather than the target of, sensorimotor integration. As sensors become more ubiquitous in everyday life, there is an opportunity to expand sensing beyond the device to directly inform the prosthesis of the current state of the user. By directly sensing the state of the individual, new possibilities emerge for the prosthesis to behave in synergy with and adjust its actions to align with those of the individual. This dissertation offers a shift in perspective from targeting purely joint-level outcomes, to instead focusing on system-level considerations toward a holistic approach to prosthetic control, considering the prosthesis and individual as one system rather than separate components. Specifically, this dissertation contributes novel advancements to control algorithms for powered lower-limb prostheses through the development of a Task-space Control (TSC) framework. The design, application, and assessment of TSC is presented as follows. First, normative task-space characteristics, namely center-of-mass (CoM) kinematics and ground reaction forces (GRFs), are accurately replicated by optimizing well-established template models of legged locomotion. The custom trajectory optimization framework revisits the constant leg stiffness assumption of these models to accurately reconstruct experimental CoM kinematics and GRFs. By varying the leg stiffness throughout the gait cycle to account for the non-energetically conservative interplay of muscle and joint coordination, the framework reconstructs CoM kinematics within the observed step-to-step variance in human walking, while also matching normative GRFs. Moreso, the resulting leg stiffness profiles align with observed trends in leg stiffness for human walking. Second, the TSC framework is presented for control of a powered ankle prosthesis for level-ground walking and compared to both a passive and a state-of-the-art continuous impedance controller across multiple speeds. Desired task-space references are generated via the first contribution, with the GRFs converted into a feedforward torque output at the ankle and CoM feedback layered on top. Without any explicit joint-level considerations, TSC generates ankle kinematics and kinetics that qualitatively resemble normative characteristics and improve ankle joint symmetry compared to the passive mode. Third, TSC is extended to knee-ankle control for level-ground walking and stair ascent. The first contribution is updated to generate desired CoM kinematics and GRF information for both gait modes. The TSC framework generates knee and ankle kinematics that qualitatively resemble normative kinematics for both level-ground and stair ascent, emphasizing the generalizability of TSC beyond single-joint, single-task implementations. User feedback focuses on whether replicating normative joint-level characteristics for the sake of joint-level characteristics is the correct path forward for prosthesis control. By shifting the perspective of prosthesis control away from purely joint-level considerations, the TSC framework illustrates the ability to achieve desired joint-level responses from system-level considerations, effectively aligning the behavior of the prosthesis to the overall movements of the individual.<p></p>