Co-evolutionary particle swarm optimization algorithm for two-sided robotic assembly line balancing problem

TLDR

Two‑sided assembly lines with robots are common for large‑volume products, yet few studies address their balancing problems. The study seeks to minimize cycle time in two‑sided robotic assembly line balancing. A mixed‑integer programming model is solved by CPLEX for small instances, and a co‑evolutionary particle swarm optimization—enhanced with local search, global‑best modification, and restart—is applied to benchmark problems. The proposed algorithm outperforms most competing metaheuristics on the majority of the seven test problems.

Abstract

Industries utilize two-sided assembly lines for producing large-sized volume products such as cars and trucks. By employing robots, industries achieve a high level of automation in the assembly process. Robots help to replace human labor and execute tasks efficiently at each workstation in the assembly line. From the literature, it is concluded that not much work has been conducted on two two-sided robotic assembly line balancing problems. This article addresses the two-sided robotic assembly line balancing problem with the objective of minimizing the cycle time. A mixed-integer programming model of the proposed problem is developed which is solved by the CPLEX solver for small-sized problems. Due to the problems in non-polynomial--hard nature, a co-evolutionary particle swarm optimization algorithm is developed to solve it. The co-evolutionary particle swarm optimization utilizes local search on the global best individual to enhance intensification, modification of global best to emphasize exploration, and restart mechanism to escape from local optima. The performances of the proposed co-evolutionary particle swarm optimization are evaluated on the modified seven well-known two-sided assembly line balancing problems available in the literature. The proposed algorithm is compared with five other well-known metaheuristics, and computational and statistical results demonstrate that the proposed co-evolutionary particle swarm optimization outperforms most of the other metaheuristics for majority of the problems considered in the study.