Hybrid Robust Multi-Objective Evolutionary Optimization Algorithm

摘要

A hybrid robust multi-objective optimization algorithm and accompanying software were developed that: 1) utilize several evolutionary optimization algorithms, a set of rules for automatic switching among these algorithms in order to accelerate the overall convergence and avoid termination in a local minimum, 2) involve development of algorithms for multi-dimensional response surfaces (metamodels) that are fast, accurate and robust by utilizing wavelet-based artificial neural networks, polynomials of radial basis functions, and multi-layer self adapting maps, 3) involve an algorithm based on Bayesian statistics (using Kalman filters and Monte Carlo Markov chains) that will enhance robustness of the multi-objective optimization algorithm by accounting for uncertainties in the input data and in the accuracy of the evaluation methods for the multiple objective functions. The hybrid evolutionary multi- objective optimization algorithm was also thoroughly tested on a number of standard test problems with two and three simultaneous objectives where the Pareto surface could be continuous and discontinuous. The hybrid optimizer was programmed in such a way that it can be transportable to any single-processor or parallel processor.