Non-linear optimization: artificial neural network solution techniques applied to the optimum linear feedback control of linear discrete-time dynamic systems

A new methodology for the solution of constrained nonlinear optimization problems is proposed. Originally grown out of the necessity for obtaining the best linear law to control linear discrete-time dynamic systems (LDTDS), it can be used in every optimization problem of both linear and non-linear cost functions and constraints. An appropriate procedure for handling both equality and inequality constraints is offered along with its application on real-world problems. A powerful artificial neural network (ANN) is implemented to fully exploit the proposed technique and experimental results are provided. The chaotic behaviour of the latter is also discussed.<
>