Time optimal control using pi-sigma networks

A controller for providing time-optimal trajectory of objects is proposed based on the pi-sigma neural network. This network facilitates the use of a priori knowledge of the problem in determining suitable weight initializations. A contracting delta descent algorithm is introduced for direct adaptive control, which avoids the necessity for an identification model and ensuing backpropagation of error. Simulation results using a controller for the actuator of a hard disk drive show a smooth trajectory with no overshoot or ringing