Multiobjective Evolutionary Algorithm Based on Dynamic Encoding and Population Isolating

A multiobjective evolutionary algorithm, suitable for complex high-dimensional multiobjective optimization problems, was proposed based on dynamic encoding and population isolating. The main ideas included: (1) dynamical encoding that could strengthen the capability of global and local search; (2) individual evaluation that helped for improving population diversity; (3) population isolating that divided an evolutionary population into different degraded subpopulations through individual fitness while each of them was evolved according to the designated crossover probability and adaptive gene segment mutation rules; (4) collection and update, i.e., an outer set and excellent individual pool with different usage were to collect the best individuals of evolving populations while being updated by a clustering algorithm. Compared to several representative multiobjective evolutionary algorithms through application to several extremely difficult optimization problems with high dimensions, the proposed algorithm shows great superiority and application potential