Human stance control beyond steady state response and inverted pendulum simplification

摘要

Systems theory analyses have suggested that human upright stance can be modelled in terms of continuous multi-sensory feedback control. So far, these analyses have considered mainly steady-state responses to periodic stimuli and relied on a simplifying model of the body’s mechanics in the form of an inverted pendulum. Therefore, they may have ignored relevant aspects of the postural behaviour. To prove a more general validity of a stance control model that we previously derived from such analyses, we now presented subjects with static–dynamic stimulus combinations and assessed response transients, anterior–posterior (a–p) response asymmetries, and possible deviations from the ‘inverted pendulum’ simplification (by measuring hip and knee bending). We presented normal subjects (Ns) and vestibular loss patients (Ps) with a–p support surface tilt on a motion platform under the instruction to maintain, with eyes closed, the body upright in space. In addition, subjects were to indicate perceived platform tilt with the help of pointers. We combined a fixed-amplitude sinusoidal tilt (0.1 Hz) with static tilts that were varied in amplitude and direction. We recorded upper body (shoulder) and lower body (hip) excursions in space and centre of pressure (COP) shift, and calculated the centre of mass (COM) angular excursion. We found that: (1) Immediately prior to stimulus onset (which was highly predictable), subjects showed a small anticipatory forward lean. (2) The subsequent transient response consisted of two parts. First, the body was moved along with the platform tilt and then, in the second part, the body excursion was braked by starting tilt compensation. Upon increasing tilt amplitude, the braking point showed a pronounced saturation with for-aft asymmetry. (3) During the following prolonged tilt, the tonic body excursions saturated with increasing static tilt amplitude. This saturation also showed a for-aft asymmetry (backwards saturation more pronounced). In contrast, the dynamic body excursions did not depend on the static tilt stimulus. (4) Tilt compensation occurred mainly in the ankle joints, but also involved small synergistic bendings in hips and knees in fixed register to the ankle rotation. (5) After the end of the stimulus, the body returned towards primary position, followed by a pronounced and slowly decaying tonic overshoot which was mainly related to tilt amplitude and initial tonic body excursion. (6) The responses of Ps qualitatively resembled those of Ns, apart from larger body excursions, less pronounced saturations, and less for-aft asymmetries. (7) Perceived platform tilt of Ns and Ps was correlated with their postural tilt compensations, but unlike the postural responses the perceptual responses overestimated actual static and dynamic tilt by a factor of 3–4. Our findings suggest two, so far undescribed and highly nonlinear mechanisms in human stance control. (a) The braking during the transient response appears to reflect a ‘sensory reweighting switch’ by which subjects change from a control that is referenced to the support to one that is referenced to space. (b) The saturation of the tonic body excursion also reflects a sensory reweighting mechanism; by this, subjects keep their balancing within a certain excursion limit. The two mechanisms were originally not predicted by our stance control model, but do not invalidate it, because they can simply be added to it. Also the observed for-aft asymmetries can be accounted for (by making thresholds in the two mechanisms asymmetric). In its extended form, the model mimics the previous and the new findings. We also conclude that the ‘inverted pendulum’ simplification is a legitimate simplification. We demonstrate the utility of the model by implementing it into a humanoid robot that then mimics closely the human experimental data. Finally, we present a hypothetical concept on sensory reweighting mechanisms in human stance control, which is meant to serve as a framework for future research.