Signal synthesis by means of evolutionary algorithms

Abstract

In this article, we investigate a procedure for generating signals with genetic algorithms. Signals are obtained from elementary patterns characterized by different degrees of freedom. These patterns are repeated and combined in order to reach specific signal shapes. The whole signal parametrization has to be determined by solving a difficult inverse problem of high dimensionality and strong multimodality. This can be carried out using evolutionary algorithms with the aim of finding all pattern configurations in the signal. The different signal synthesis schemes are evaluated, tested and applied to the generation of particular railway driving profiles.