Investigation of a Unified Phase Variable in Human Locomotion for Applications in Powered Prosthese

Estimates indicate that by 2050 the U.S. will incur a two-fold increase in the incidence of amputation and stroke, due largely to the prevalence of vascular disease. Amputees suffer from a slower, less stable, and less efficient gait than that of able-bodied persons. Developing methods to control powered prosthetic legs in a simple, efficient, and customizable manner could help current and future amputees ambulate more efficiently. The current methodology used to control powered prosthetic legs sees the gait cycle as a process consisting of discrete states (e.g., heel strike, load acceptance, toe off, pre-swing, mid-swing, etc). Thus, current powered prosthetic legs synchronize to their wearer by transitioning between a finite numbers of states based on switching rules. This approach is limited as state machine control strategies end up dictating the walking speed and response of the robotic leg rather than the amputee. Novel approaches from the biped robotic field have led to new ways of visualizing the gait cycle. The gait cycle is seen and controlled as a continuous periodic process synchronized by a phase variable rather than a sequence of discrete events. A phase variable is a mechanical signal that changes monotonically, i.e., it strictly increases or decreases, over time and therefore is able to parameterize a rhythmic process. This approach allows biped robots to synchronize all their joint actuators seamlessly in order to achieve a stable and steady gait. In this dissertation, this type of control architecture is translated to control a powered prosthetic leg by studying the human gait cycle from a phase variable perspective. Before designing a controller for a powered prosthetic leg that is capable of matching the timing of the amputee's gait cycle, the human sense of phase in locomotion must be understood. In this dissertation, the design of a perturbation mechanism and experimental protocol capable of producing multi-joint phase-shifting perturbations in able-bodied subjects is presented. These phase-shifting perturbations helped us statistically compare different parameterizations of nominal and perturbed joint trajectories in ten able-bodied subjects. We derived and analyzed unified phase variable candidates (i.e., able to parameterize the entire stride) in human locomotion from the results of these perturbation experiments. A statistical analysis found the correlations between nominal and perturbed joint trajectories to be significantly greater when parameterized by the derived unified phase variable (0.95+) than by time. Finally, we performed experiments on three above knee amputee subjects wearing a powered knee and ankle prosthetic leg with a unified phase variable that correctly parameterizes perturbed kinematics and is suitable for real-time controllers. Implementing our bioinspired phase variable in the robotic leg not only allowed amputee subjects to walk steadily at different speeds and inclines, but also gave them voluntary control over the leg during non-rhythmic tasks (e.g., walking backwards and stepping back and forth). In addition, no tuning of the controller was required between subjects due to the versatility of the phase variable algorithm. The results from the amputee experiments show that we have developed a control architecture that allows the powered prosthetic leg to behave as a plug-and-play device due to the simple, yet effective, derivation of a unified phase variable.