Neural networks and digital avionics

The application of neural networks is considered as a method of solution to a number of outstanding problems in aircraft avionics. The areas of application of artificial neural networks to avionics dealt with are: target selection, attack planning/steering, and track-before-detect. The target selection is approached by the application of a feedforward, backpropagation network. The attach planning/steering is approached by a novel type of parallel processing neural network. The track-before-detect is solved via a feedforward backpropagation network. The feedforward backpropagation algorithms can be implemented on fast systolic-array-type neural chips. A special, fast path generation chip can be developed. Consequently, a low-cost, high-speed, compact solution to a number of avionics functions is available through neural networks.<>