Design and Model-Based Safety Verification of a Volitional Sit-Stand Controller for a Powered Knee-Ankle Prosthesis

ABSTRACT Sit-stand transitions, the motions executed by individuals to stand up or sit down, are an important determinant of overall mobility and a common source of falls. Unilateral amputees using standard passive prostheses are further challenged by sit-stand transitions due to muscle and joint asymmetries they exhibit between the sound

and amputated sides, often resulting in debilitating back pain. Powered knee-ankle prostheses can produce enough torque to assist meaningfully during sit-stand transitions and can meet design criteria such as producing smooth motion on the amputated side that matches the sound side. Controllers for these prostheses

can be designed to allow user-driven control of the leg. However, the production of high torques not directly commanded by the user comes with increased risks. This is of particular concern because these legs must be adopted outside of controlled lab environments. Thus, any powered prosthesis must demonstrably meet design

and safety criteria. While safety-critical medical devices, such as pacemakers, are subjected to extensive testing and validation procedures, there is no agreed-upon standard in the powered prosthetics field for how to define and measure safety. Prior work on sit-stand controllers has focused only on measuring a limited number

of outcomes with respect to one design criterion on a small number of subjects, providing no guarantees about safety. The set of techniques known as formal verification provides powerful tools to reason about the behavior of systems that are composed of interacting mechanical, software, and biological modules. Given a model of a

system, formal verification allows us to probe the system’s behavior over an infinite range of possibilities that cannot be replicated in the lab during a typical testing session. These methods can then guide real-world testing, and alert system designers to problematic regions of execution. In this project, I propose to apply

formal verification techniques to design a volitional controller for sit-stand transitions with provable safety guarantees, using physics-based models and novel mathematical formulations of safety. The University of Michigan Robotics Institute is one of the top institutes of its kind in the US and provides an

ideal environment and infrastructure for the successful completion of this research. The Robotics Institute gait lab has all of the necessary equipment needed for powered prosthesis research, including two state-of-the-art prosthetic legs, and access to advanced computational resources such as the Great Lakes high performance

computing cluster. Drs. Gregg and Ozay have proven expertise relevant to the aims of this project, and will provide mentorship that will guide my research, my training, and the attainment of my career goals.