The Inverse Problem in Cognitive, Perceptual and Proprioceptive Control of Sensorimotor Behaviour Towards a Biologically Plausible Model of the Control of Aiming Movements.

Aiming movements with an arm (one to three joints) confined in a vertical plane are described in terms of control theory. The framework includes two levels of motor control, namely the discrete and the continuos level, both with respect to time. On the continuos level, movement control is configured for physical execution. On the discrete level, planning takes place, which covers the selection of a perceptual goal, its serialisation into a desired trajectory by pattern generation, the perceptual paralleling of the resulting actual movement trajectory by pattern recognition, and the final check whether the goal is hit. The fundamental term is the ‘tool transformation’, which quantitatively relates muscularly generated joint torques to the effects on the environment – here to the changing Cartesian positions of the arm’s tip. Feedforward control then is exerted by an inverse model of the tool transformation. The acquisition of the inverse model is guarantied by the proposed ‘learner-operator model ’ of sensorimotor control, which is based on auto-imitation combined with Hebbian learning, masters redundant degrees of freedom, and allows to quickly update the model parameters while the arm is moving. The approach relinquishes of a forward model of the tool transformation, and explains anticipation as predisposed by the perceptive goal, not as predicted by efferences flowing through a forward model