A review of neural-fuzzy controllers for robotic manipulators

Abstract

This paper presents a literature search on the development of three main approaches-namely neural networks, fuzzy logic and a combination of neural networks and fuzzy logic (neural-fuzzy)-for the intelligent control of robotic manipulators. The conventional computed torque method is first reviewed and its disadvantages highlighted. Several schemes using neural networks are then presented and compared. The characteristics of using neural networks are summarised. Next, the paper reviews and compares the features, strengths and weaknesses of three schemes of fuzzy logic controllers. The common drawbacks of using fuzzy logic are also highlighted. Finally, an approach which fuses fuzzy logic and neural networks is discussed.