Sensor-Driven Musculoskeletal Dynamics Modeling

Abstract

While there exist a number of frameworks that model the human musculoskeletal system, we cannot yet predict detailed human arm dynamics during manipulation tasks. A system that can predict the forces involved would revolutionize our ability to create safe and effective exoskeletons and prostheses and would allow for new and extensive study of human motor control. However, the limitations of non-invasive sensing and the wide range of morphological differences between individuals has hampered the creation of an adequate model. In this talk, I will discuss two projects underway in the Human Assistive Robotic Technologies (HART) Lab: 1) an attempt to create a predictive dynamics model of the human arm using a wide range of sensing modalities, including ultrasound, surface EMG, and acoustic myography, and 2) a framework to evaluate the quality of these models using high-resolution MRI data from multiple subjects. Together, these projects represent progress toward human musculoskeletal models that are fully customizable to individuals, regardless of injury or pathology.