Analog Neuro-Fuzzy network for system modeling and control

The aim of this work is to present a new Neuro-Fuzzy network based on a neural model called Approximate Identity Neural Network. The ability of neural networks to learn from examples and the attitude of fuzzy model to code human knowledge can be helpfully joined to create and adaptive fuzzy system. Our architecture is particularly suited to be implemented by analogue CMOS VLSI hardware. The small dimension required and the semplicity of interfacing analogue hardware with sensors make this network eligible for low cost embedded application. The architecture, a circuit implementation and a simple implementations example with SPICE simulation are presented.