PROJECT ABSTRACT The overall goal of this project is to model human joint biomechanics over continuously-varying locomotion to enable adaptive control of powered above-knee prostheses. Above-knee amputees often struggle to perform the varying activities of daily life with conventional prostheses due to the lack of positive mechanical work and active control. Emerging powered prostheses have motors that can perform these missing functions, but the biomechanics experienced by the user depend on the control of these motors. The way the prosthesis interacts with both the user and environment can be controlled through joint impedance—the relationship between joint motion and torque. Prosthetic joint impedance is typically defined via a stiffness, viscosity, and equilibrium angle for discrete phases of gait within a limited set of discrete activities, but this framework does not allow continuous variations of steady-state activities (e.g., walking at different speeds/inclines) or continuous transitions between activities (e.g., walk to stair ascent). The central hypothesis of this project is that variable joint impedance can be parameterized by a continuous model based on measurable quantities called phase and task variables. This project will use machine learning to identify variable impedance functions from able-bodied data including joint perturbation responses across the phase/task space to bias the solution toward biological values. The resulting impedance model will be used with real-time estimates of phase and task variables to control a custom powered knee-ankle prosthesis and the Össur PowerKnee across activities. The specific aims are as follows: Aim 1: Model and control joint impedance over continuums of steady-state activities. Hypothesis: We hypothesize that knee and ankle impedance during walking and stair climbing can be modeled and controlled as continuous functions of gait phase, speed, and terrain inclination. Aim 2: Model and control joint impedance over continuous transitions between activities. Hypothesis: We hypothesize that knee and ankle impedance during continuous transitions between sitting, standing, walking, and stair climbing can be modeled and controlled as functions of transitional task variables. Aim 3: Implement continuous impedance controller and clinician interface on commercial hardware. Hypothesis: The impedance control system implemented on the Össur PowerKnee with a partially-automated tuning process will improve amputee endurance and symmetry over activities of daily life. This project will be scientifically significant to understanding how human joint mechanics continuously adapt to varying tasks and terrains, technologically significant to designing agile, biomimetic powered prosthetic legs for community ambulation, and clinically significant to the widespread adoption of intuitive powered prostheses for improved patient outcomes. The adaptability and clinical viability of this control paradigm will have a pro...