Nonlinear MISO modeling using genetic programming

An algebraic model for an unknown, nonlinear, MISO (multiple input, single output) system is derived from a table of the system's input and output values. Genetic programming is used to find a model that is optimal (or nearly optimal) with respect to a nonlinear performance index. In order to apply genetic programming to this task, an encoding strategy to represent the model is devised. Then, specialized genetic operators are defined to refine the solution. The technique is shown to produce a good model for a simple nonlinear example having two inputs and one output.