Contribution to design of complex mechatronic systems. An approach through evolutionary optimization

This paper describes a new evolutionary methodology aimed at optimizing various and heterogeneous data in common evolution. The representation of solutions uses mixed-integer genotypes and variable-length chromosomes to face a complex problem of task decomposition and high-level control generation. A memory operator is introduced to face convergency uncertainties issued from the irregularities of both discontinuous evaluation function and heterogeneous solution representation. The stability of the evolutionary algorithm is analyzed with dimension changes in the optimization problem.