A modified average-roulette cellular automaton algorithm for optimization tasks.

The majority–minority cellular automaton algorithm (MmCAA) is a novel optimization algorithm that has demonstrated its competitiveness compared to other algorithms. However, one of the main issues with MmCAA is the global exploration capability. It affects the convergence speed and the ability of the algorithm to solve problems in higher dimensions. To tackle this issue, a modified average-roulette cellular automaton algorithm (ARCAA) is proposed in this paper. ARCAA comprises two core rules: the average rule and the roulette rule. The combination of the two rules contributes to the well-rounded evolution of cellular states. In the individual cell exploitation phase, the cell state undergoes random and targeted disturbances and evolves in different directions. In the single neighbor exploration phase, the cell state is guided by reference values of randomly selected neighbors and adjusted in the appropriate directions. The performance of the proposed ARCAA is evaluated by 26 benchmark functions and three real-world engineering problems. It is compared with state-of-the-art optimization algorithms. Meanwhile, the significance of the comparisons is verified through the Wilcoxon signed-rank test. These results indicate that ARCAA effectively balances exploration and exploitation, showing strong potential for broader application in various optimization problems. [ABSTRACT FROM AUTHOR]