Title: Control of standing balance described by computational reinforcement learning model
Thesis Supervisor: Dr. Jean-Sébastien Blouin
Co-Supervisor: Dr. Calvin Kuo
Committee Members: Dr. Patrick Forbes, Dr. Michiel Van De Panne
Chair: Dr. Mark Carpenter
Abstract: Standing balance is a complex sensorimotor integration process and is also important for everyday activities, including walking, stair-climbing, and sports. Prior studies have explored computational models of balance control using deterministic controllers and explained the control rules of standing balance by studying the outcome of the control action in the world. However, all those models require internal variables to record properties of the dynamic of standing balance. This assumption is not proven since the brain receives information and generates commands without necessarily knowing about the dynamic of movement. With this thesis proposal, I aim to get one step closer to a physiological computational model by stochastically modeling standing balance control along the Anterior-Posterior (AP) direction. The control rule of the model will not be based on internal representations, however, it will be obtained by the interaction with a dynamic model of standing balance as the human body. Here I will not model physiological sensory information and physiological motor commands, which could be considered a future work after completing this thesis. Moreover, the deterministic controllers have failed to model exploratory actions and the interaction of these actions with the cost of energy and mechanical constraints. Furthermore, since previous models inherently had a linear nature, they cannot explain the nonlinearity of the control in standing balance. Here, by the stochastic model of standing balance, I also aim to explore how humans adopt their preferred positions based on the mechanical boundaries, decode control gains of the response to angle perturbations based on posture state in standing balance, and update physiological computational models of human standing balance control. This thesis proposal might inspire new engineering applications, such as AI-powered tools to combine multiple sensors to solve complex problems.