Human Posture Control on a Dynamic Platform

Abstract

This thesis presents a comprehensive investigation into human postural control strategies when subjected to underfloor perturbations representing real-world scenarios. This study diverges from traditional postural control research, which primarily investigates standing balance on platforms with large amplitudes (0.1-0.2 m). Instead, it explores balance strategies on a platform moving just 4mm in the anterior-posterior direction. The study was carried out in three phases.

In the initial phase, we conducted experimental studies involving nine healthy participants. Standing balance was examined on a platform whose movement frequency was systematically scaled from 0.4 to 6 Hz. Employing motion capture and force plate systems, we recorded platform motion as well as kinematic and kinetic information of our participants. A vector coding approach quantified coordination between hip, knee, and ankle joint torques, and the centre of mass (COM) and centre of pressure (COP) motion. Our experimental results revealed a significant main effect of platform frequency for knee-ankle and COP-COM phase relationships. A transition from the ankle to knee strategy was observed at certain frequencies, signifying the emergence of knee strategy as a novel mechanism in real-world settings.

In the second phase, we developed a Model Predictive Control (MPC) approach to human balance control in response to these underfloor perturbations. This model not only simulates the complexity of the human body's control system but also provides predictive capabilities for human postural responses. Our MPC-based scheme effectively captured several intricate aspects of human balance control system, including the accurate prediction of stability loss, transition between in-phase to anti-phase COP-COM relative motion, and autonomous selection of balance control strategies (ankle and knee).

In the final phase, we incorporated sensory-motor noise into our MPC model to assess its robustness and performance under realistic scenarios. By considering two major sources of noise—process noise during movement execution and measurement noise related to sensory feedback—we could simulate a more authentic sensory-motor system. We employed a "state observer" to estimate state variables and compensate for noise, thereby enhancing the predictive model's effectiveness despite the inherent uncertainty in sensory-motor signals.

Together, these findings not only advance our understanding of the complex interplay between different joint strategies in maintaining standing balance but also contribute to the development of more accurate and robust predictive models in the field of human postural control.