DESIGN OF FININTE STATE MACHINES EMPLOYING BACK PROPAGATION IN ARTIFICIAL NEURAL NETWORKS

Abstract

Design of Finite State machines has a lot of applications ranging from the simple vending machines to complex robotics for space exploration. Recently the need for anticipative state machines exhibiting Artificial Intelligence has come up. In this paper, anticipative state machines have been designed using Artificial Neural Networks. Back propagation in ANN has been resorted to with the aim of attaining low errors in prediction and lesser training time, both of which are critical parameters for real time applications. It has been shown that the proposed technique renders lesser prediction errors with respect to increasing number of inputs as compared to previously existing techniques.