Five advanced multiobjective optimization algorithms—adaptive geometry estimation-based multiobjective evolutionary algorithm (AGE-MOEA) and its enhanced version, adaptive grid-based evolutionary multiobjective algorithm II (AGE-MOEA-II), non-dominant sorting genetic algorithm II (NSGA-II), non-dominant sorting genetic algorithm III (NSGA-III) and multiple objective particle swarm optimization (MOPSO) —are employed to determine the optimal layout strategies for multiple tower cranes. The goal is to minimize safety risks while maximizing coverage rate. Additionally, the performance of these five optimization algorithms is compared.
Based on previous research, this paper considers the number, radius and positioning of multiple tower cranes as decisive variables while incorporating temporary facilities as constraints. Five advanced multiobjective optimization algorithms—AGE-MOEA, AGE-MOEA-II, NSGA-II, NSGA-III and MOPSO—are used to identify optimal tower crane layout strategies, aiming to minimize safety risks and maximize coverage rate. Subsequently, these algorithms are evaluated through a performance comparison using generation distance (GD), generation distance plus (GD+), inverted generation distance (IGD), inverted generation distance plus (IGD+) and hypervolume (HV).
Research confirms the effectiveness of these five algorithms in optimizing multiple tower cranes layout. The results show that MOPSO produces lower quality pareto front solutions than the other four algorithms, with a significant gap between optimization outcomes for safety risks and coverage rate. While MOPSO has the shortest runtime, its performance is suboptimal. Each algorithm has its strengths, and the final choice should be based on project-specific requirements.
This study employs five algorithms—AGE-MOEA, AGE-MOEA-II, NSGA-II, NSGA-III and MOPSO—to optimize the layout of multiple tower cranes, balancing safety risks minimization with maximum coverage rate. It also laid the foundation for optimizing the layout of more complex multiple tower cranes. Our analysis, combining graphical visualizations with rigorous performance metrics, provides data-driven support for evaluating each algorithm’s strengths and limitations.
