Training a fuzzy controller to back up an autonomous vehicle

A design method for a fuzzy controller based on emulating an existing controller is presented. In this method, the human operator is not required to provide a set of explicit linguistic fuzzy control rules. Training information can be obtained by observing the input-output behavior of the human operator, where the target controller output for a given input is the control action of the human operator for that input. Thus the fuzzy controller learns to copy the control strategies of the human operator. Simulations show that a fuzzy controller can be trained to steer an autonomous vehicle while backing up to a loading dock. The technique presented should be applicable to a wide variety of nonlinear control problems.<
>