A neurological view on reactive human stance control ★

During biped stance or locomotion, humans show remarkable skills in reactive balancing upon external disturbances. Mainly four types of external disturbances are relevant for stance: the field force gravity, contact forces such as a push or pull against the body, as well as body support surface rotation and translational acceleration. It is known from clinics that sensory loss severely impairs the balancing. Three sensory inputs are instrumental: vestibular, joint angle, and joint torque. System identification studies currently try to understand how humans are able to flexibly cope with changes in, and superposition of the disturbances. A solution is presented in this article. The article first describes the control problem and then reviews recent evidence for a PD (proportional-derivative) controller, for multisensory feedback, and for sensory reweighting as a key to understand the flexibility. On this basis, a recent disturbance estimation and compensation (DEC) model is introduced. It builds on two concepts from previous psychophysical studies of human self-motion perception. First, inputs from several sensory transducers are fused to establish sensors that provide explicit measures of the physical variables (sensor concept). Second, a processing level, interleaved between sensors and feedback (meta level), performs online sensory estimations of the external disturbances. These estimations are then fed into a local proprioceptive feedback loop, yielding corresponding disturbance rejections. Previous work on using the DEC model to describe human reactive balancing data is briefly reviewed. Then, novel work is presented, in which voluntary control over the reactive balancing is added to the DEC model. A prediction method for anticipating self-produced and external disturbances is suggested and corresponding software and hardware (robot) simulations are presented. The results serve as guidelines for future human experiments. Since the DEC model is very simple, we concluded that its behavioral flexibility and fault tolerance goes together with computational parsimony, an equally important biological constraint.