NEUROPT — Optimal Control Framework for Studying the Neural Control of Force, Impedance, and Feedback Responses

A hallmark of human motor behavior is our ability to control movements during physical interaction with the environment. Yet our understanding of the neuromechanical processes that enable the control of physical interaction is still severely limited. Our limited understanding of how humans control physical interaction is a significant limitation for many applications, especially in the field of human-robot interaction, as modeling the human response to external inputs would allow researchers to optimize the action of wearable active systems in silico at least in part, for example by identifying rehabilitation strategies that intervene on specific neuromuscular pathways linked to a control policy of interest. Experimental evidence indicates that humans will control physical interaction using a combination of three strategies: 1) update of the force input, 2) upregulation of state feedback responses to rapidly respond to deviations between the predicted and received sensory input; 3) impedance control, based on coactivation of agonist and antagonist muscle pairs. Unfortunately, no single theory of motor control is currently able to simultaneously account for all of these control policies, or to link them to specific processes in the central nervous system. We have assembled a team of investigators with a unique combination of computational and experimental skills and capabilities that we propose to combine to develop a new theory of the neuromuscular control of interaction suitable for studying the optimal control of force, impedance, and feedback responses. Specifically, PI Berret, an expert in computational models of motor control, has developed an optimal control framework suitable for studying the force and impedance control; PI Cashaback, an expert in computational modeling of muscle mechanics, has developed a computationally tractable muscle model describing muscle force and impedance; PI Sergi, an expert in human-robot interaction and instrumentation, has developed a family of MRI-compatible wrist robots able to operate alongside functional magnetic resonance imaging (fMRI) for studying the neural control of motor tasks involving physical interaction. In this project, we propose to combine our efforts to develop a new computational theory to understand the optimal control of force, impedance, and feedback responses, for human movements in presence of physical interaction with the environment. We will validate and refine the computational theory via behavioral experiments conducted in the lab involving exposure to dynamic distortion conditions with optimal characteristics as identified by model-based optimization, and replicated during fMRI to identify the neural correlates of the input and states of the model.