ABSTRACT
Presented is the solution to a nonlinear missile control problem through application of the backpropagation algorithm to a feed-forward artificial neural network that uses memory neurons. Memory neuron networks have recently shown promise in identification and adaptive control as the memory neurons allow the network to combine historical values with present backpropagation weights and outputs [1,2]. The learning algorithm employs only locally available information rather than feeding some or all past plant inputs to the network. This solves the memory problem associated with storing all past values of the missile/target system in order to apply a new control to minimize the difference error.
