Neural adaptive predictor for visual tracking system
F. Lunghi, Stefano Lazzari, Giovanni Magenes · 2002
The present work introduces a neural adaptive predictor, which can be inserted in classical models of human visual tracking system (made up by the combination of the Saccadic and Smooth Pursuit systems), in order to explain and simulate humans ability to compensate the 130 ms physiological delay when they follow the movement of an external target with the eyes. The neural predictor, previously trained to accomplish the task, can improve his performance modifying on-line his parameters (weights). The parameter changes rely on a cost function obtained by statistical evaluation of the positional error, measured at the output of the system, i.e. after a transmission delay. Although this algorithm has been developed for the eye tracking system, it has not been tailored on it and it can be applied on a great variety of tracking systems with delays in the forward path or in the actuators.