Falls are the leading cause of fatal and nonfatal injuries in elderly people, resulting in approximately $31 billion in medical costs annually in the U.S. These injuries motivate balance control studies focused on improving stability by identifying prevention strategies for reducing the number of fall events. Experiments provide data about subjects’ kinematic response to loss of balance. However, simulations offer additional insights, and may be used to make predictions about functional outcomes of interventions. Several approaches already exist in biomechanics research to generate accurate models on a subject-by-subject basis. However, these representations typically lack models of the central nervous system, which provides essential feedback that humans use to make decisions and alter movements. Interdisciplinary methods that merge biomechanics with other fields of study may be the solution to fill this gap by developing models that accurately reflect human neuromechanics.

Roboticists have developed control systems approaches for humanoid robots simultaneously accomplishing complex goals by coordinating component tasks under priority constraints. Concepts such as the zero-moment point and extrapolated center of mass have been thoroughly evaluated and are commonly used in the design and execution of dynamic robotic systems in order to maintain stability. These established techniques can benefit biomechanical simulations by replacing biological sensory feedback that is unavailable in the virtual environment. Subject-specific simulations can be generated by synthesizing techniques from both robotics and biomechanics and by creating comprehensive models of task-level coordination, including neurofeedback, of movement patterns from experimental data.

In this work, we demonstrate how models built on robotic principles that emulate decision making in response to feedback can be trained by biomechanical motion capture data to produce a subject-specific fit. The resulting surrogate can predict a subject’s particular solution to accomplishing the movement goal of recovering balance by controlling component tasks. This research advances biomechanics simulations as we move closer towards the development of a tool capable of anticipating the results of rehabilitation interventions aimed at correcting movement disorders. The novel platform presented here marks the first step towards that goal, and may benefit engineers, researchers, and clinicians interested in balance control and falls in human subjects.