Comparative Study of Non Linear System Modeling Using Artificial Intelligent Techniques

Abstract

Models of real system are of fundamental importance in virtually all disciplines. The models are useful for system analysis i.e., or gaining a better understanding of the system. Models make it possible to predict or simulate a system’s behavior. In engineering, models are required for the design of new processes and for the analysis of existing processes. Advanced techniques for the design of controllers, optimizations, supervision, and fault detection are also based on process model. In this paper a simulated comparison of fuzzy logic and neural network control of the truck backer-upper system is presented. The aim of the controller is to back a truck to a loading dock which is a difficult task. It is a nonlinear control problem for which no traditional control system design method exists. We assumed that there were no linguistic rules available, and therefore the controllers were designed from the available numerical data only. We provided the same desired input-output pairs to both the neural and the fuzzy approaches, and compared the final control performance of both models. It is known that artificial neural networks and fuzzy logics are powerful tools for handling problems of large dimension. Many studies have been reported on the ability of neural networks and fuzzy logics for approximating nonlinear functions. The chores of this paper are to model the truck backer upper control problem using neural networks and fuzzy logic. This paper proposes a comparative study of neural networks (FFN, RBFN and RNN) and fuzzy logic for modeling the truck backer upper control problem. The body angle ø, x position and the steering angle θ of the truck are used as training data for neural network and fuzzy logic. The results showed the superiority of the neural controller over the fuzzy one, when the later was influenced by the amount of overlapping between its sets and the missing rules from its rule base.