Quantized Controller with Close to Optimum Nonlinear Control Surface

The paper investigates a new kind of controller--the quantized controller, which can give a close to optimum nonlinear control surface. In contrary to the rough control surface generated by the fuzzy controller, the quantized controller can give better and smoother control. The quantized controller is trainable. By using an error function and the gradient decent method, control surface can be fine-tuned to some close to optimum control surface. The classical example of the backing truck to ramp problem is investigated using both the fuzzy controller and the quantized controller. Results show that the trained quantized controller can perform faster and smoother operations than the fuzzy controller.