A parallel computing architecture for intelligent sensory data processing

It is noted that enhanced sensing capabilities in advanced systems of automation place an increasing computational burden on the control system architecture. Sensory processing for intelligent systems will take a variety of forms, ranging from simple preprocessing of sensory input values to sophisticated fusion and decision analysis algorithms. Sensory preprocessing, particularly when based upon state estimation methods, can take maximal advantage of a variety of parallel array processing techniques, while more advanced forms of decision analysis based upon sensory data can use special-purpose hardware units to optimize the largely serial throughput of these algorithms. A multifunctional sensory processing (MFSP) architecture is proposed for meeting some of these diverse requirements. The elements and applications of the MFSP architecture are discussed.<<ETX>>