Multiplayer Battle Game-Inspired Optimizer for Complex Optimization Problems

TL;DR

The statistical analysis results reveal that the novel MBGO demonstrates significant competitiveness, excelling in convergence speed and achieving high levels of convergence accuracy across both benchmark functions and real-world problems.

Abstract

Various popular multiplayer battle royale games share a lot of common elements. Drawing from our observations, we summarized these shared characteristics and subsequently proposed a novel heuristic algorithm named multiplayer battle game-inspired optimizer (MBGO). The proposed MBGO streamlines mainstream multiplayer battle royale games into two discrete phases: movement and battle. Specifically, the movement phase incorporates the principles of commonly encountered ``safe zones'' to incentivize participants to relocate to areas with a higher survival potential. The battle phase simulates a range of strategies adopted by players in various situations to enhance the diversity of the population. To evaluate and analyze the performance of the proposed MBGO, we executed it alongside eight other algorithms, including three classics and five latest ones, across multiple diverse dimensions within the CEC2017 and CEC2020 benchmark functions. In addition, we employed several industrial design problems to evaluate the scalability and practicality of the proposed MBGO. The results of the statistical analysis reveal that the novel MBGO demonstrates significant competitiveness, excelling not only in convergence speed, but also in achieving high levels of convergence accuracy across both benchmark functions and real-world problems.

Figures (9)

• Figure 1: The universal optimization framework of proposed MBGO algorithm.
• Figure 2: Screenshots from two different games, where the entire map is divided into two different areas.
• Figure 3: The average convergence curves of all competitor algorithms on 50-D CEC2020 functions.The horizontal coordinate represents the total number of evolutions of the current individual and the vertical coordinate represents the fitness value of the current optimal solution.
• Figure 4: The average convergence curves of all competitor algorithms on 100-D CEC2020 functions. The horizontal coordinate and the vertical coordinate are the same as Fig. \ref{['CEC2020_50D']}.
• Figure 5: The mean and standard deviation of the optimal solutions obtained from 30 trial runs for all algorithms are reported for the 10-D CEC2017 function functions. In this context, symbols $+$ and $-$ respectively signify that the proposed algorithm outperforms or underperforms the comparison algorithm. The symbol $=$ denotes no significant difference between the two. Furthermore, for each function, the optimal solutions identified among all algorithms are highlighted in bold.The last row of the table represents the "number of functions that outperform the comparison algorithm / number of functions that are significantly different / number of functions that are lower than the comparison algorithm".