A neuroengineering solution to the optimal tracking problem

A mathematical solution is developed for an optimal sensory-to-motor transformation. This specifies a unique vector of motor command signals for goal-directed upper limb movement, conditional on cost of muscular effort vs. performance accuracy. Derivation is based on a realistic model of visual tracking, incorporating characteristics of external tracking system, target and disturbance as well as multivariable, nonlinear, time-varying characteristics of neuromuscular and biomechanical systems internal to the human operator. The optimal transformation removes redundancy in a 58-dimensional muscle system to give a two-dimensional response, thus solving the degrees-of-freedom problem. While adaptive filter neural networks are required to implement the general solution, an instructive linear matrix approximation reveals computational modules with observable behavioural correlates such as prediction, synergy generation and speed-accuracy compromise. (C) 1999 Elsevier Science B.V. All rights reserved.