An estimation of distribution algorithm with branch-and-bound based knowledge for robotic assembly line balancing

Robotic assembly lines are widely used in manufacturing industries. The robotic assembly line balancing (RALB) problem aims to balance the workloads among different workstations and optimize the assembly line efficiency. This paper addresses a particular type of RALB problem, which minimizes the assembly line cycle time by determining the task and robot assignment in each workstation under precedence constraints. To solve the problem, we present an effective hybrid algorithm fusing the estimation of distribution algorithm and branch-and-bound (B&B) based knowledge. A problem-specific probability model is designed to describe the probabilities of each task being assigned to different workstations. Based on the probability model, an incremental learning method is developed and a sampling mechanism with B&B based knowledge is proposed to generate new feasible solutions. The fuse of B&B based knowledge is able to reduce the search space of EDA while focusing the search on the promising area. To enhance the exploitation ability, a problem-specific local search is developed based on the critical workstation to further improve the quality of elite solutions. The computational complexity of the proposed algorithm is analyzed, and the effectiveness of the B&B based knowledge and the problem-specific local search is demonstrated through numerical experiments. Moreover, the performance of the proposed algorithm is compared with existing algorithms on a set of widely-used benchmark instances. Comparative results demonstrate the effectiveness and efficiency of the proposed algorithm.