GSoC/GCI Archive
Google Summer of Code 2013 PaGMO / PyGMO

Constraints Handling Techniques in Evolutionary Algorithms

by Jérémie Labroquère for PaGMO / PyGMO

Because of the stochastic behavior of Evolutionary Algorithms, some constraints handling methods use empirical parameters defined by the user to take into account constraints violation during the optimization process. Unfortunately these parameters can drive the optimizer to local optima and are usually effective for a single problem. This proposal consists in implementing in PyGMO/PagMO more or less sophisticated constraints handling techniques that are problem independent and to test them against a set of typical problems. In an implementation point of view, most of the constraint handling techniques can be seen as a modification of the initial constrained problem, algorithm or population. Thus, the implementation will consists in redefining the initial problem/algorithm/population through the use of operators/modifiers.