Hybrid Intelligent Approaches in Robotics

Abstract

Although fuzzy logic can encode expert knowledge in a direct and easy way using rules with linguistic labels, it often takes a lot of time to design and tune the membership functions which quantitatively define these linguistic labels. Wrong membership functions can lead to poor controller performance and possible instability. An excellent solution is to apply learning techniques by neural networks, which can be used to design membership functions automatically, simultaneously reducing development time and costs and improving the system performance. These combined neuro-fuzzy networks can learn faster than neural networks. They also provide a connectionist architecture that is easy for VLSI implementation to perform the functions of a conventional fuzzy logic controller with distributed learning abilities.