Integrating neural networks, fuzzy logic and genetic algorithms for intelligent control

In this thesis, the integration of Neural Networks (NNs), Fuzzy Logic (FL) and Genetic Algorithms (GAs) for intelligent control is proposed and applied to the classical problem of docking a truck.

A backpropagation neural network architecture using a "step" update weight modification is used to obtain quickly and efficiently trajectory data from given initial states. A new algorithm to define fuzzy logic rules is used on the trajectory data to build a fuzzy logic knowledge base. This fuzzy logic knowledge base is then optimised using a genetic algorithm to obtain a fuzzy logic controller that effectively simulates a full neural network solution to the problem of docking of a truck.