A Neural Network Parallel Adaptive Controller for Dynamic System Control

Abstract

A neural network (NN)-based intelligent adaptive controller that introduces a new concept of intelligent supervisory loop is proposed. The scheme consists of an online radial basis-function NN (RBFNN) in parallel with a model reference adaptive controller (MRAC) and uses a growing dynamic RBFNN to augment MRAC. Updating of the RBFNN width, center, and weight characteristics is performed such that error reduction and improved tracking accuracy are accomplished. The RBFNN architecture adapts its centers and radii and tunes the relevant parameters dynamically. These adaptations effectively address the issues that are related to initial error and dimensional growth that are inherent in static NN design. The strength of the proposed scheme is in its ability to perform effectively, even when the plant mode swings and functional changes occur. Theoretical results are validated by simulation studies based on a nonlinear single-link flexible robotic manipulator position tracking of changing reference pattern. Compared to single and multiple fuzzy reference adaptive control approaches, the proposed intelligent controller produced better tracking with reduced tracking error in the event of functional changes and is capable of delivering plant output to track the reference precisely.