Adaptive behavior design based on FNN for the mobile robot

A fuzzy neural network (FNN) has been trained off-line to memory the fuzzy control rules of the adaptive behaviors for the local optimal path planning of mobile robot. The fuzzy rules were collected automatically by reinforcement Q_Learning (QL) on-line beforehand. This method has overcome the disadvantage of traditional means which are determined by artificial experience, and are able to meet the requirement of local optimal path planning. The FNN controller is constructed by BP neural network (BPNN). After the off-line training, the rules are stored implicitly in FNN. In control applications, without looking up table, when the real-time sense information is input to FNN, the best adaptive behavior is produced in the output. The simulation results show that because all training samples are from the trained fuzzy rules, the output is almost the same as the result of the training rules.