Abstract
After the model of artificial neuron was proposed in 1943, artificial neural network has become an important tool for learning, but not creation. In this dissertation, we simulate human's learning and creating abilities to a control system, and propose "A Control System based on Continuous Learning and Creation (CSCLC)". It is constructed by fuzzy neural network and algorithms. CSCLC has two modules: Learning module using numerical data to construct fuzzy control rule, and to learn membership function by fuzzy neural network. Creating module contains three sub-parts: After each control process, performance index sub-part provides a performance index. If the index is not desirable, then creation is motivated. When creation motivation is generated, goal creation sub-part will search the desire control output. Then rule control sub-part will transfer it to fuzzy control rules. We have compared the proposed system with other two methods using "Backing up a Truck". The result shows that three control systems are comparable, but only CSCLC can successfully back up a truck in some initial conditions. This is because that CSCLC can adaptive different control environment by creating new control rules.