E. Miranda

An HW/SW co-design approach for neuro-fuzzy hardware design

Artificial Neural Networks (NNs) and Fuzzy Systems (FS) are highly parallel structures that consist of a large number of elementary nonlinear units (called neurons) fully interconnected. In this paper we present a HW/SW co-design approach to the partition of a complex control architecture on a digital hardware implementation of NNs combined with a host computer (DSP systems, PCs, etc.) resulting in a powerful system used in different applications. The NN processes most of the tasks, while the host performs signal pre-processing and learning algorithms. In some control applications the host also tracks some discrete states of the plant by implementing a finite state automata and/or verifying plant safety boundaries operations. The tight link with the host allows the NN hardware to be very simple since several operations related with the NNs (learning, weights initialization etc.) can be performed by the host computer. The board we built using these techniques, can be cased to implement intelligent control paradigms mixing neuro-fuzzy algorithms with finite state automata and/or digital control algorithms.

A CPWM Synapsis for Weighted Radial Basis Functions

This paper describes a silicon synapsis designed to implement Weighted Radial Basis Functions. The synapsis is based on Pulse Stream computation principles, which offer interesting performance, especially for what power dissipation and computation speed concerns. Weighted Radial Basis Functions integrate the advantages of Multi-Layer Perceptrons and Radial Basis Functions alone, therefore the silicon neural networks which results may find applications in several pattern recognition and classification tasks, especially in low power environments. Furthermore it can also be used as a method to map Fuzzy Inference Systems on silicon Artificial Neural Networks.

Hybrid neuro-fuzzy system for control of complex plants

Artificial neural networks (ANNs) and fuzzy systems (FS) are high parallel structures that consist of a large number of elementary nonlinear units (called neurons) fully interconnected. In this article we present a hardware implementation of ANNs combined with a DSP resulting in a powerful system used in control applications. The NN processes most of the control tasks, while the DSP performs signal pre-processing and learning algorithms. In some cases the DSP also tracks some discrete states of the plant by implementing a finite state automata and/or verifying plant safety boundaries operations. The tight link with a DSP allows the NN hardware to be very simple since several operations related with the NNs (learning, weights refresh, etc.) can be performed by the DSP. Moreover the system can implement intelligent control paradigms mixing neuro-fuzzy algorithms with finite state automata and/or digital control algorithms.