A neuro-fuzzy core for DSP-based controls of complex systems

Artificial neural networks (NNs) are highly parallel structures that consist of a large number of elementary nonlinear units (called neurons) fully interconnected. We present a hardware implementation of NNs combined with a DSP; both NNs and DSP combined together result in a powerful system used in control applications. NNs process most of the control tasks, while DSP performs signal preprocessing, signal conditioning and learning algorithms. In some cases the DSP performs the control of some discrete states of the plant by implementing finite state automata and/or verifying plant safety boundary operations. With this "marriage", NN hardware can be very simple because several operations related with the NNs (learning algorithms, weight maintenance, etc.) can be performed by the DSP. The system implements intelligent control paradigms by mixing neuro fuzzy algorithms with finite state automata and/or digital control algorithms.