Identifying Plant and Feedback in Human Posture Control

Thumbnail Image


Publication or External Link






Human upright bipedal stance is a classic example of a control system consisting of a plant (i.e., the physical body and its actuators) and feedback (i.e., neural control) operating continuously in a closed loop. Determining the mechanistic basis of behavior in a closed loop control system is problematic because experimental manipulations or deficits due to trauma/injury influence all parts of the loop. Moreover, experimental techniques to open the loop (e.g., isolate the plant) are not viable because bipedal upright stance is not possible without feedback. The goal of the proposed study is to use a technique called closed loop system identification (CLSI) to investigate properties of the plant and feedback separately.

Human upright stance has typically been approximated as a single-joint inverted pendulum, simplifying not only the control of a multi-linked body but also how sensory information is processed relative to body dynamics. However, a recent study showed that a single-joint approximation is inadequate. Trunk and leg segments are in-phase at frequencies below 1 Hz of body sway and simultaneously anti-phase at frequencies above 1 Hz during quiet stance. My dissertation studies have investigated the coordination between the leg and trunk segments and how sensory information is processed relative to that coordination. For example, additional sensory information provided through visual or light touch information led to a change of the in-phase pattern but not the anti-phase pattern, indicating that the anti-phase pattern may not be neurally controlled, but more a function of biomechanical properties of a two-segment body. In a subsequent study, I probed whether an internal model of the body processes visual information relative to a single or double-linked body. The results suggested a simple control strategy that processes sensory information relative to a single-joint internal model providing further evidence that the anti-phase pattern is biomechanically driven.

These studies suggest potential mechanisms but cannot rule out alternative hypotheses because the source of behavioral changes can be attributed to properties of the plant and/or feedback. Here I adopt the CLSI approach using perturbations to probe separate processes within the postural control loop. Mechanical perturbations introduce sway as an input to the feedback, which in turn generates muscle activity as an output. Visual perturbations elicit muscle activity (a motor command) as an input to the plant, which then triggers body sway as an output. Mappings of muscle activity to body sway and body sway to muscle activity are used to identify properties of the plant and feedback, respectively. The results suggest that feedback compensates for the low-pass properties of the plant, except at higher frequencies. An optimal control model minimizing the amount of muscle activation suggests that the mechanism underlying this lack of compensation may be due to an uncompensated time delay. These techniques have the potential for more precise identification of the source of deficits in the postural control loop, leading to improved rehabilitation techniques and treatment of balance deficits, which currently contributes to 40% of nursing home admissions and costs the US health care system over $20B per year.