Sensorimotor control of human dynamic behavior in space implemented into a hominoid robot

Abstract

To what extent can we claim nowadays that we understand sensorimotor control of human dynamic behavior in space? We try here to answer this question by exploring whether the available knowledge base suffices to build a hominoid robot such that its sensorimotor control functions mimic those of humans. It is, actually, our aim to build such a robot. We want to use it, in a systems approach, for simulations to better understand human sensorimotor control functions. We posit that a systems approach is necessary to deal with this complex non-linear control. We are especially interested in the sensory aspects of the control, the inter-sensory interactions ('multisensory integration' or sensor fusion) and the spatio-temporal coordination. Psychophysical work in our laboratory showed that the brain creates from sensory inputs internal estimates of the physical stimuli in the outside world (i.e., of the external constellation that caused a particular set of sensor stimuli). For example, the brain derives from vestibular and proprioceptive signals an estimate of body support surface motion. It then uses these estimates for sensorimotor feedback control (rather than the 'raw' sensory signals such as the vestibular signal). We hold that this internal reconstruction of the external physics is required for appropriate spatio-temporal coordination of the behavior. However, a problem arises from non-ideal sensors. An example is the vestibular sensor, which shows pronounced low-frequency noise. The solution of this problem involves sensory re-weighting mechanisms. Based on the discovered sensor fusion principles, we built a hominoid robot for control of upright stance (which we consider a simple prototype of sensorimotor control). It mimics human stance control even in complex behavioral situations. We aim to use it to better understand sensorimotor deficits in neurological patients and to develop new therapy designs.