This research was motivated by the need to improve the efficiency of the Genetic Algorithm
(GA) when dealing with a variety of complex problems. The goal is to develop strategies that
allow the GA to automatically adjust itself to the specific challenges of each problem, without
the need for manual intervention to readjust its operational parameters, making this algorithm
a more dynamic tool. To achieve this goal, this research proposed two bioinspired strategies to
enhance the adaptability and efficiency of GA. The first strategy was Adaptive Radiation (AR),
a biological phenomenon that causes high rates of mutation in populations, allowing rapid
adaptation to survival conditions. The second strategy was a selection technique inspired by
Multi-Criteria Decision Models (MCDM) and the natural behavior of choosing partners,
observed on different species, which assist in decision making, evaluating solutions based on
multiple criteria. The methodology consists of implementing these strategies in GA, creating
two new algorithms: GA with Adaptive Radiation (GAAR) and Multicriteria GA (MCGA).
These algorithms were then tested on three different categories of problems: ten benchmark
functions, which simulate a variety of complex environments; four engineering problems,
which represent industry challenges; and a real problem, to test the practical applicability of
the algorithms in a high magnitude scenario. The results showed that the GAAR and MCGA
algorithms outperformed the standard GA and other optimization algorithms on most of the
tested problems. In particular, they were able to effectively adapt to different types of problems
and find efficient solutions without the need to manually readjust their parameters. These
results suggest that the introduction of bioinspired strategies such as AR and MCDM can
significantly improve GA performance, making them a powerful tool for a wide range of realworld applications.
|
