Following natural evolution closely in devising evolutionary algorithms for certain challenging problems

Since their inception, evolutionary algorithms have been presented as computational realisations of natural evolution in a grossly macro-philosophical sense. Concepts such as survival of the fittest, recombination, mutation, and elite preservation are all inspired by observations of biological evolution and genetics, and are broadly implemented to capture their essential principles. Despite their enormous success over the past four decades and their extensions to solve various types of search and optimisation problems, there still exist certain challenging problems where the above-mentioned simplistic extensions are not enough. We present one such problem, which, in addition to the final optimal solution, demands the discovery of a sequence of transitional solutions to transform the currently implemented solution to the final optimal solution. A systematic solution approach to such problems is highly practical and enables a user with a sequence of resource-restraint solutions of small changes to arrive at the target solution. The required gradual nature of a sequence of transitional solutions forces us to follow nature's gradualism principle and modify the original problem into a bi-objective formulation and modify a usual evolutionary multi-objective optimisation algorithm into a more effective nature-inspired procedure. Results on a keyboard layout optimisation and a numerical optimisation problem are shown for illustration purpose. We demonstrate that existing EMO algorithms which do not utilise gradualism fail to solve such challenging problems, while the proposed algorithm emphasises intermediate and gradual solutions temporarily through generations in an automated manner for solving such problems. This study not only supports the mimicry of nature's evolutionary principle in addressing search and optimisation problems, but also highlights the importance of introducing further details of the natural evolutionary process to tackle more complex and demanding search and optimisation problems.