Distributed Neural Network for Smart Structures Applications

Neural control algorithms have been the subject of intensive study and development over the last ten years. Their chief advantages are their potntial to control many sensors and actuators in a distributed fashion, thus relieving the burden levied on each computational unit; and their ability to optimize their own performance based on measured data, reducing or eliminating their dependence on highly questionable and complex models of the system to be controlled. Both of these advantages are crucial to the success of future missions, but their realization depends absolutely on the existence of a neural network that is low-cost and efficient, both in terms of its impact on system resources such as weight, size and power and in terms of development costs.Planning Systems Inc. (PSI) proposes an innovative approach which leverages COTS local area network technology to provide a fas, modular neural network. The proposed system will enable algorithm designers to develop and implement high spped algorithms in an efficient, high-level software environment.