Virtual trajectory and stiffness ellipse during force-trajectory control using a parallel-hierarchical neural network model

Proposes a parallel-hierarchical neural network model using a feedback-error-learning scheme. This model explains the biological motor learning for simultaneous control of both trajectory and force. Moreover, the authors propose a control law based on a criterion related to the minimum motor-command-change trajectory. The motor commands are calculated while directly taking account of variable viscous-elastic properties of muscles. Learning trajectory and force control is performed for a two-link four-muscle arm. They derive the virtual trajectory and stiffness ellipse, which are implicitly determined during force and trajectory control. They found that the virtual trajectory was much more complex than the desired trajectory. The stiffness ellipses were similar to those obtained in Mussa-Ivaldi experiment.<>