Central Target Pattern Generation, Neural Feedforward and Mechanical Negative Feedback in Motion Control: How Muscles Drive Voluntary Movements While Maintaining Posture Against Gravitation

K. T. Kalveram, S. Beirle, S. Richter, P. Jansen, & J. Konczak

University of Duesseldorf, Duesseldorf, Germany


The present paper suggests "hybrid motor controller" which combines three types of control strategies mostly treated seperately, namely: Central target pattern generation, here referring to the acceleration of the intended movement, feedforward control (here neurally mediated and using "inverse modelling of the tool transformation" of the plant to be controlled), and negative feedback control (here realized exclusively through the mechanical spring properties of the limb-muscle system). Neural feedforward by inverse modelling is used to compensate for gravitational and/or other predictable forces, thus enabling—also in a feedforward fashion—a simple realization of the time course put out by the central target generator. Thereby, mechanical negative feedback control via the springlike properties of the limb/muscle system is employed only if the inverse model is not correct, or an unforeseen external perturbation acts upon the limb chain. This spring can, if necessary, operate with high stiffness (that is to say, with a high loop gain), but without loosing stability. The reasons are, that the mechanical system exhibits no time delays, and that the myotatic reflex activity, if given, only modulates—with a comparatively long time constant—the parameter of stiffness, but does not influence the control signal immediately. Therefore, the proposed "hybrid motor controller" contrasts to alpha-type and to lambda-type models of motor control as well.

The controller model has been tested by comparing real lower arm movements—with and without perturbations—to simulated arm movements under the same conditions and with the same physical parameters. Referring to the time courses of position, velocity and acceleration, there exist striking similarities between the kinematics of the real and the artificial movements. Therefore, the hybrid control model outlined above has a good chance to reflect natural control modes of the body as well.